def get_symbol(args, arg_params, aux_params):
data_shape = (args.image_channel,args.image_h,args.image_w)
image_shape = ",".join([str(x) for x in data_shape])
if args.network[0]=='d':
embedding = fdensenet.get_symbol(512, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='m':
print('init mobilenet', args.num_layers)
embedding = fmobilenet.get_symbol(512,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(512,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(512,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(512, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
else:
print('init resnet', args.num_layers)
embedding = fresnet.get_symbol(512, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
gt_label = mx.symbol.Variable('softmax_label')
assert args.loss_type>=0
extra_loss = None
if args.loss_type==0:
_weight = mx.symbol.Variable('fc7_weight')
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding, weight = _weight, bias = _bias, num_hidden=args.num_classes, name='fc7')
elif args.loss_type==1:
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, 512), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
fc7 = mx.sym.LSoftmax(data=embedding, label=gt_label, num_hidden=args.num_classes,
weight = _weight,
beta=args.beta, margin=args.margin, scale=args.scale,
beta_min=args.beta_min, verbose=1000, name='fc7')
elif args.loss_type==2:
_weight = mx.symbol.Variable('fc7_weight')
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding, weight = _weight, bias = _bias, num_hidden=args.num_classes, name='fc7')
print('center-loss', args.center_alpha, args.center_scale)
extra_loss = mx.symbol.Custom(data=embedding, label=gt_label, name='center_loss', op_type='centerloss',\
num_class=args.num_classes, alpha=args.center_alpha, scale=args.center_scale, batchsize=args.per_batch_size)
elif args.loss_type==10: #marginal loss
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')
params = [1.2, 0.3, 1.0]
n1 = mx.sym.expand_dims(nembedding, axis=1) #N,1,C
n2 = mx.sym.expand_dims(nembedding, axis=0) #1,N,C
body = mx.sym.broadcast_sub(n1, n2) #N,N,C
body = body * body
body = mx.sym.sum(body, axis=2) # N,N
#body = mx.sym.sqrt(body)
body = body - params[0]
mask = mx.sym.Variable('extra')
body = body*mask
body = body+params[1]
#body = mx.sym.maximum(body, 0.0)
body = mx.symbol.Activation(data=body, act_type='relu')
body = mx.sym.sum(body)
body = body/(args.per_batch_size*args.per_batch_size-args.per_batch_size)
extra_loss = mx.symbol.MakeLoss(body, grad_scale=params[2])
elif args.loss_type==11: #npair loss
params = [0.9, 0.2]
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')
nembedding = mx.sym.transpose(nembedding)
nembedding = mx.symbol.reshape(nembedding, (512, args.per_identities, args.images_per_identity))
nembedding = mx.sym.transpose(nembedding, axes=(2,1,0)) #2*id*512
#nembedding = mx.symbol.reshape(nembedding, (512, args.images_per_identity, args.per_identities))
#nembedding = mx.sym.transpose(nembedding, axes=(1,2,0)) #2*id*512
n1 = mx.symbol.slice_axis(nembedding, axis=0, begin=0, end=1)
n2 = mx.symbol.slice_axis(nembedding, axis=0, begin=1, end=2)
#n1 = []
#n2 = []
#for i in xrange(args.per_identities):
# _n1 = mx.symbol.slice_axis(nembedding, axis=0, begin=2*i, end=2*i+1)
# _n2 = mx.symbol.slice_axis(nembedding, axis=0, begin=2*i+1, end=2*i+2)
# n1.append(_n1)
# n2.append(_n2)
#n1 = mx.sym.concat(*n1, dim=0)
#n2 = mx.sym.concat(*n2, dim=0)
#rembeddings = mx.symbol.reshape(nembedding, (args.images_per_identity, args.per_identities, 512))
#n1 = mx.symbol.slice_axis(rembeddings, axis=0, begin=0, end=1)
#n2 = mx.symbol.slice_axis(rembeddings, axis=0, begin=1, end=2)
n1 = mx.symbol.reshape(n1, (args.per_identities, 512))
n2 = mx.symbol.reshape(n2, (args.per_identities, 512))
cosine_matrix = mx.symbol.dot(lhs=n1, rhs=n2, transpose_b = True) #id*id, id=N of N-pair
data_extra = mx.sym.Variable('extra')
data_extra = mx.sym.slice_axis(data_extra, axis=0, begin=0, end=args.per_identities)
mask = cosine_matrix * data_extra
#body = mx.sym.mean(mask)
fii = mx.sym.sum_axis(mask, axis=1)
fij_fii = mx.sym.broadcast_sub(cosine_matrix, fii)
fij_fii = mx.sym.exp(fij_fii)
row = mx.sym.sum_axis(fij_fii, axis=1)
row = mx.sym.log(row)
body = mx.sym.mean(row)
extra_loss = mx.sym.MakeLoss(body)
elif args.loss_type==12:
_weight = mx.symbol.Variable('fc7_weight')
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding, weight = _weight, bias = _bias, num_hidden=args.num_classes, name='fc7')
params = [0.9, 0.2]
nembedding = mx.symbol.slice_axis(embedding, axis=0, begin=0, end=args.images_per_identity)
nembedding = mx.symbol.L2Normalization(nembedding, mode='instance', name='fc1n')
n1 = mx.sym.expand_dims(nembedding, axis=1)
n2 = mx.sym.expand_dims(nembedding, axis=0)
body = mx.sym.broadcast_sub(n1, n2) #N,N,C
body = body * body
body = mx.sym.sum(body, axis=2) # N,N
body = body - params[0]
body = mx.symbol.Activation(data=body, act_type='relu')
body = mx.sym.sum(body)
n = args.images_per_identity
body = body/(n*n-n)
extra_loss = mx.symbol.MakeLoss(body, grad_scale=params[1])
#extra_loss = None
else:
#embedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*float(args.loss_type)
embedding = embedding * 5
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, 512), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance') * 2
fc7 = mx.sym.LSoftmax(data=embedding, label=gt_label, num_hidden=args.num_classes,
weight = _weight,
beta=args.beta, margin=args.margin, scale=args.scale,
beta_min=args.beta_min, verbose=100, name='fc7')
#fc7 = mx.sym.Custom(data=embedding, label=gt_label, weight=_weight, num_hidden=args.num_classes,
# beta=args.beta, margin=args.margin, scale=args.scale,
# op_type='ASoftmax', name='fc7')
if args.loss_type<=1 and args.incay>0.0:
params = [1.e-10]
sel = mx.symbol.argmax(data = fc7, axis=1)
sel = (sel==gt_label)
norm = embedding*embedding
norm = mx.symbol.sum(norm, axis=1)
norm = norm+params[0]
feature_incay = sel/norm
feature_incay = mx.symbol.mean(feature_incay) * args.incay
extra_loss = mx.symbol.MakeLoss(feature_incay)
#out = softmax
#l2_embedding = mx.symbol.L2Normalization(embedding)
#ce = mx.symbol.softmax_cross_entropy(fc7, gt_label, name='softmax_ce')/args.per_batch_size
#out = mx.symbol.Group([mx.symbol.BlockGrad(embedding), softmax, mx.symbol.BlockGrad(ce)])
out_list = [mx.symbol.BlockGrad(embedding)]
softmax = None
if args.loss_type<10:
softmax = mx.symbol.SoftmaxOutput(data=fc7, label = gt_label, name='softmax', normalization='valid')
out_list.append(softmax)
if softmax is None:
out_list.append(mx.sym.BlockGrad(gt_label))
if extra_loss is not None:
out_list.append(extra_loss)
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params):
data_shape = (args.image_channel, args.image_h, args.image_w)
image_shape = ",".join([str(x) for x in data_shape])
margin_symbols = []
if args.network[0] == 'd':
embedding = fdensenet.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'm':
print('init mobilenet', args.num_layers)
if args.num_layers == 1:
embedding = fmobilenet.get_symbol(
args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
elif args.network[0] == 'i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(
args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0] == 's':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
elif args.network[0] == 'y':
print('init mobilefacenet', args.num_layers)
embedding = fmobilefacenet.get_symbol(
args.emb_size,
bn_mom=args.bn_mom,
version_output=args.version_output)
else:
print('init resnet', args.num_layers)
embedding = fresnet.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit,
version_act=args.version_act)
all_label = mx.symbol.Variable('softmax_label')
gt_label = all_label
extra_loss = None
_weight = mx.symbol.Variable("fc7_weight",
shape=(args.num_classes, args.emb_size),
lr_mult=args.fc7_lr_mult,
wd_mult=args.fc7_wd_mult)
if args.loss_type == 0: #softmax
if args.fc7_no_bias:
fc7 = mx.sym.FullyConnected(data=embedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
else:
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding,
weight=_weight,
bias=_bias,
num_hidden=args.num_classes,
name='fc7')
elif args.loss_type == 1: #sphere
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
fc7 = mx.sym.LSoftmax(data=embedding,
label=gt_label,
num_hidden=args.num_classes,
weight=_weight,
beta=args.beta,
margin=args.margin,
scale=args.scale,
beta_min=args.beta_min,
verbose=1000,
name='fc7')
elif args.loss_type == 2:
s = args.margin_s
m = args.margin_m
assert (s > 0.0)
assert (m > 0.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
s_m = s * m
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=s_m,
off_value=0.0)
fc7 = fc7 - gt_one_hot
elif args.loss_type == 4:
s = args.margin_s
m = args.margin_m
assert s > 0.0
assert m >= 0.0
assert m < (math.pi / 2)
_weight = mx.symbol.Variable("fc7_weight",
shape=(args.num_classes, args.emb_size),
lr_mult=1.0)
if args.finetune:
print("{}finetuning from trained model".format('-' * 10))
_weight = mx.symbol.Variable("finetune_weight",
shape=(args.num_classes,
args.emb_size),
lr_mult=10.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
cos_m = math.cos(m)
sin_m = math.sin(m)
mm = math.sin(math.pi - m) * m
#threshold = 0.0
threshold = math.cos(math.pi - m)
if args.easy_margin:
cond = mx.symbol.Activation(data=cos_t, act_type='relu')
else:
cond_v = cos_t - threshold
cond = mx.symbol.Activation(data=cond_v, act_type='relu')
body = cos_t * cos_t
body = 1.0 - body
sin_t = mx.sym.sqrt(body)
new_zy = cos_t * cos_m
b = sin_t * sin_m
new_zy = new_zy - b
new_zy = new_zy * s
if args.easy_margin:
zy_keep = zy
else:
zy_keep = zy - s * mm
new_zy = mx.sym.where(cond, new_zy, zy_keep)
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=1.0,
off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
elif args.loss_type == 5:
s = args.margin_s
m = args.margin_m
assert s > 0.0
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
if args.margin_a != 1.0 or args.margin_m != 0.0 or args.margin_b != 0.0:
if args.margin_a == 1.0 and args.margin_m == 0.0:
s_m = s * args.margin_b
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=s_m,
off_value=0.0)
fc7 = fc7 - gt_one_hot
else:
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
t = mx.sym.arccos(cos_t)
if args.margin_a != 1.0:
t = t * args.margin_a
if args.margin_m > 0.0:
t = t + args.margin_m
body = mx.sym.cos(t)
if args.margin_b > 0.0:
body = body - args.margin_b
new_zy = body * s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=1.0,
off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
out_list = [mx.symbol.BlockGrad(embedding)]
softmax = mx.symbol.SoftmaxOutput(data=fc7,
label=gt_label,
name='softmax',
normalization='valid')
out_list.append(softmax)
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params):
data_shape = (args.image_channel, args.image_h, args.image_w)
image_shape = ",".join([str(x) for x in data_shape])
margin_symbols = []
if args.network[0] == 'd':
embedding = fdensenet.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'm':
print('init mobilenet', args.num_layers)
if args.num_layers == 1:
embedding = fmobilenet.get_symbol(
args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
elif args.network[0] == 'i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(
args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0] == 's':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
else:
print('init resnet', args.num_layers)
embedding = fresnet.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
all_label = mx.symbol.Variable('softmax_label')
if not args.output_c2c:
gt_label = all_label
else:
gt_label = mx.symbol.slice_axis(all_label, axis=1, begin=0, end=1)
gt_label = mx.symbol.reshape(gt_label, (args.per_batch_size, ))
c2c_label = mx.symbol.slice_axis(all_label, axis=1, begin=1, end=2)
c2c_label = mx.symbol.reshape(c2c_label, (args.per_batch_size, ))
assert args.loss_type >= 0
extra_loss = None
if args.loss_type == 0: #softmax
_weight = mx.symbol.Variable('fc7_weight')
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding,
weight=_weight,
bias=_bias,
num_hidden=args.num_classes,
name='fc7')
elif args.loss_type == 1: #sphere
_weight = mx.symbol.Variable("fc7_weight",
shape=(args.num_classes, args.emb_size),
lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
fc7 = mx.sym.LSoftmax(data=embedding,
label=gt_label,
num_hidden=args.num_classes,
weight=_weight,
beta=args.beta,
margin=args.margin,
scale=args.scale,
beta_min=args.beta_min,
verbose=1000,
name='fc7')
elif args.loss_type == 8: #centerloss, TODO
_weight = mx.symbol.Variable('fc7_weight')
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding,
weight=_weight,
bias=_bias,
num_hidden=args.num_classes,
name='fc7')
print('center-loss', args.center_alpha, args.center_scale)
extra_loss = mx.symbol.Custom(data=embedding, label=gt_label, name='center_loss', op_type='centerloss',\
num_class=args.num_classes, alpha=args.center_alpha, scale=args.center_scale, batchsize=args.per_batch_size)
elif args.loss_type == 2:
s = args.margin_s
m = args.margin_m
_weight = mx.symbol.Variable("fc7_weight",
shape=(args.num_classes, args.emb_size),
lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
if s > 0.0:
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
if m > 0.0:
if args.margin_verbose > 0:
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
margin_symbols.append(mx.symbol.mean(cos_t))
s_m = s * m
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=s_m,
off_value=0.0)
fc7 = fc7 - gt_one_hot
if args.margin_verbose > 0:
new_zy = mx.sym.pick(fc7, gt_label, axis=1)
new_cos_t = new_zy / s
margin_symbols.append(mx.symbol.mean(new_cos_t))
else:
fc7 = mx.sym.FullyConnected(data=embedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
if m > 0.0:
body = embedding * embedding
body = mx.sym.sum_axis(body, axis=1, keepdims=True)
body = mx.sym.sqrt(body)
body = body * m
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=1.0,
off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, body)
fc7 = fc7 - body
elif args.loss_type == 3:
s = args.margin_s
m = args.margin_m
assert args.margin == 2 or args.margin == 4
_weight = mx.symbol.Variable("fc7_weight",
shape=(args.num_classes, args.emb_size),
lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
if args.margin_verbose > 0:
margin_symbols.append(mx.symbol.mean(cos_t))
#threshold = math.cos(args.margin_m)
#cond_v = cos_t - threshold
#cond = mx.symbol.Activation(data=cond_v, act_type='relu')
#body = cos_t
#for i in xrange(args.margin//2):
# body = body*body
# body = body*2-1
#new_zy = body*s
#zy_keep = zy
#new_zy = mx.sym.where(cond, new_zy, zy_keep)
#if args.margin_verbose>0:
# new_cos_t = new_zy/s
# margin_symbols.append(mx.symbol.mean(new_cos_t))
#diff = new_zy - zy
#diff = mx.sym.expand_dims(diff, 1)
#gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
#body = mx.sym.broadcast_mul(gt_one_hot, diff)
#fc7 = fc7+body
elif args.loss_type == 4:
s = args.margin_s
m = args.margin_m
assert s > 0.0
assert m >= 0.0
assert m < (math.pi / 2)
_weight = mx.symbol.Variable("fc7_weight",
shape=(args.num_classes, args.emb_size),
lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
if args.output_c2c == 0:
cos_m = math.cos(m)
sin_m = math.sin(m)
mm = math.sin(math.pi - m) * m
#threshold = 0.0
threshold = math.cos(math.pi - m)
if args.margin_verbose > 0:
margin_symbols.append(mx.symbol.mean(cos_t))
if args.easy_margin:
cond = mx.symbol.Activation(data=cos_t, act_type='relu')
else:
cond_v = cos_t - threshold
cond = mx.symbol.Activation(data=cond_v, act_type='relu')
body = cos_t * cos_t
body = 1.0 - body
sin_t = mx.sym.sqrt(body)
new_zy = cos_t * cos_m
b = sin_t * sin_m
new_zy = new_zy - b
new_zy = new_zy * s
if args.easy_margin:
zy_keep = zy
else:
zy_keep = zy - s * mm
new_zy = mx.sym.where(cond, new_zy, zy_keep)
else:
#set c2c as cosm^2 in data.py
cos_m = mx.sym.sqrt(c2c_label)
sin_m = 1.0 - c2c_label
sin_m = mx.sym.sqrt(sin_m)
body = cos_t * cos_t
body = 1.0 - body
sin_t = mx.sym.sqrt(body)
new_zy = cos_t * cos_m
b = sin_t * sin_m
new_zy = new_zy - b
new_zy = new_zy * s
if args.margin_verbose > 0:
new_cos_t = new_zy / s
margin_symbols.append(mx.symbol.mean(new_cos_t))
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=1.0,
off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
elif args.loss_type == 10: #marginal loss
nembedding = mx.symbol.L2Normalization(embedding,
mode='instance',
name='fc1n')
params = [1.2, 0.3, 1.0]
n1 = mx.sym.expand_dims(nembedding, axis=1) #N,1,C
n2 = mx.sym.expand_dims(nembedding, axis=0) #1,N,C
body = mx.sym.broadcast_sub(n1, n2) #N,N,C
body = body * body
body = mx.sym.sum(body, axis=2) # N,N
#body = mx.sym.sqrt(body)
body = body - params[0]
mask = mx.sym.Variable('extra')
body = body * mask
body = body + params[1]
#body = mx.sym.maximum(body, 0.0)
body = mx.symbol.Activation(data=body, act_type='relu')
body = mx.sym.sum(body)
body = body / (args.per_batch_size * args.per_batch_size -
args.per_batch_size)
extra_loss = mx.symbol.MakeLoss(body, grad_scale=params[2])
elif args.loss_type == 11: #npair loss
params = [0.9, 0.2]
nembedding = mx.symbol.L2Normalization(embedding,
mode='instance',
name='fc1n')
nembedding = mx.sym.transpose(nembedding)
nembedding = mx.symbol.reshape(
nembedding,
(args.emb_size, args.per_identities, args.images_per_identity))
nembedding = mx.sym.transpose(nembedding, axes=(2, 1, 0)) #2*id*512
#nembedding = mx.symbol.reshape(nembedding, (args.emb_size, args.images_per_identity, args.per_identities))
#nembedding = mx.sym.transpose(nembedding, axes=(1,2,0)) #2*id*512
n1 = mx.symbol.slice_axis(nembedding, axis=0, begin=0, end=1)
n2 = mx.symbol.slice_axis(nembedding, axis=0, begin=1, end=2)
#n1 = []
#n2 = []
#for i in xrange(args.per_identities):
# _n1 = mx.symbol.slice_axis(nembedding, axis=0, begin=2*i, end=2*i+1)
# _n2 = mx.symbol.slice_axis(nembedding, axis=0, begin=2*i+1, end=2*i+2)
# n1.append(_n1)
# n2.append(_n2)
#n1 = mx.sym.concat(*n1, dim=0)
#n2 = mx.sym.concat(*n2, dim=0)
#rembeddings = mx.symbol.reshape(nembedding, (args.images_per_identity, args.per_identities, 512))
#n1 = mx.symbol.slice_axis(rembeddings, axis=0, begin=0, end=1)
#n2 = mx.symbol.slice_axis(rembeddings, axis=0, begin=1, end=2)
n1 = mx.symbol.reshape(n1, (args.per_identities, args.emb_size))
n2 = mx.symbol.reshape(n2, (args.per_identities, args.emb_size))
cosine_matrix = mx.symbol.dot(lhs=n1, rhs=n2,
transpose_b=True) #id*id, id=N of N-pair
data_extra = mx.sym.Variable('extra')
data_extra = mx.sym.slice_axis(data_extra,
axis=0,
begin=0,
end=args.per_identities)
mask = cosine_matrix * data_extra
#body = mx.sym.mean(mask)
fii = mx.sym.sum_axis(mask, axis=1)
fij_fii = mx.sym.broadcast_sub(cosine_matrix, fii)
fij_fii = mx.sym.exp(fij_fii)
row = mx.sym.sum_axis(fij_fii, axis=1)
row = mx.sym.log(row)
body = mx.sym.mean(row)
extra_loss = mx.sym.MakeLoss(body)
elif args.loss_type == 12: #triplet loss
nembedding = mx.symbol.L2Normalization(embedding,
mode='instance',
name='fc1n')
anchor = mx.symbol.slice_axis(nembedding,
axis=0,
begin=0,
end=args.per_batch_size // 3)
positive = mx.symbol.slice_axis(nembedding,
axis=0,
begin=args.per_batch_size // 3,
end=2 * args.per_batch_size // 3)
negative = mx.symbol.slice_axis(nembedding,
axis=0,
begin=2 * args.per_batch_size // 3,
end=args.per_batch_size)
ap = anchor - positive
an = anchor - negative
ap = ap * ap
an = an * an
ap = mx.symbol.sum(ap, axis=1, keepdims=1) #(T,1)
an = mx.symbol.sum(an, axis=1, keepdims=1) #(T,1)
triplet_loss = mx.symbol.Activation(data=(ap - an +
args.triplet_alpha),
act_type='relu')
triplet_loss = mx.symbol.mean(triplet_loss)
#triplet_loss = mx.symbol.sum(triplet_loss)/(args.per_batch_size//3)
extra_loss = mx.symbol.MakeLoss(triplet_loss)
elif args.loss_type == 9: #coco loss
centroids = []
for i in xrange(args.per_identities):
xs = mx.symbol.slice_axis(embedding,
axis=0,
begin=i * args.images_per_identity,
end=(i + 1) * args.images_per_identity)
mean = mx.symbol.mean(xs, axis=0, keepdims=True)
mean = mx.symbol.L2Normalization(mean, mode='instance')
centroids.append(mean)
centroids = mx.symbol.concat(*centroids, dim=0)
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * args.coco_scale
fc7 = mx.symbol.dot(nembedding, centroids,
transpose_b=True) #(batchsize, per_identities)
#extra_loss = mx.symbol.softmax_cross_entropy(fc7, gt_label, name='softmax_ce')/args.per_batch_size
#extra_loss = mx.symbol.BlockGrad(extra_loss)
else:
#embedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*float(args.loss_type)
embedding = embedding * 5
_weight = mx.symbol.Variable("fc7_weight",
shape=(args.num_classes, args.emb_size),
lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance') * 2
fc7 = mx.sym.LSoftmax(data=embedding,
label=gt_label,
num_hidden=args.num_classes,
weight=_weight,
beta=args.beta,
margin=args.margin,
scale=args.scale,
beta_min=args.beta_min,
verbose=100,
name='fc7')
#fc7 = mx.sym.Custom(data=embedding, label=gt_label, weight=_weight, num_hidden=args.num_classes,
# beta=args.beta, margin=args.margin, scale=args.scale,
# op_type='ASoftmax', name='fc7')
if args.loss_type <= 1 and args.incay > 0.0:
params = [1.e-10]
sel = mx.symbol.argmax(data=fc7, axis=1)
sel = (sel == gt_label)
norm = embedding * embedding
norm = mx.symbol.sum(norm, axis=1)
norm = norm + params[0]
feature_incay = sel / norm
feature_incay = mx.symbol.mean(feature_incay) * args.incay
extra_loss = mx.symbol.MakeLoss(feature_incay)
#out = softmax
#l2_embedding = mx.symbol.L2Normalization(embedding)
#ce = mx.symbol.softmax_cross_entropy(fc7, gt_label, name='softmax_ce')/args.per_batch_size
#out = mx.symbol.Group([mx.symbol.BlockGrad(embedding), softmax, mx.symbol.BlockGrad(ce)])
out_list = [mx.symbol.BlockGrad(embedding)]
softmax = None
if args.loss_type < 10:
softmax = mx.symbol.SoftmaxOutput(data=fc7,
label=gt_label,
name='softmax',
normalization='valid')
out_list.append(softmax)
if softmax is None:
out_list.append(mx.sym.BlockGrad(gt_label))
if extra_loss is not None:
out_list.append(extra_loss)
for _sym in margin_symbols:
_sym = mx.sym.BlockGrad(_sym)
out_list.append(_sym)
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params):
data_shape = (args.image_channel, args.image_h, args.image_w)
image_shape = ",".join([str(x) for x in data_shape])
margin_symbols = []
if args.network[0] == 'd':
embedding = fdensenet.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'm':
print('init mobilenet', args.num_layers)
if args.num_layers == 1:
embedding = fmobilenet.get_symbol(
args.emb_size,
version_input=args.version_input,
version_output=args.version_output,
version_multiplier=args.version_multiplier)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
elif args.network[0] == 'i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(
args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0] == 's':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
elif args.network[0] == 'y':
print('init mobilefacenet', args.num_layers)
embedding = fmobilefacenet.get_symbol(
args.emb_size,
bn_mom=args.bn_mom,
version_output=args.version_output)
else:
print('init resnet', args.num_layers)
embedding = fresnet.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit,
version_act=args.version_act)
all_label = mx.symbol.Variable('softmax_label')
gt_label = all_label
extra_loss = None
_weight = mx.symbol.Variable("fc7_weight",
shape=(args.num_classes, args.emb_size),
lr_mult=args.fc7_lr_mult,
wd_mult=args.fc7_wd_mult)
if args.loss_type == 0: #softmax
if args.fc7_no_bias:
fc7 = mx.sym.FullyConnected(data=embedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
else:
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding,
weight=_weight,
bias=_bias,
num_hidden=args.num_classes,
name='fc7')
elif args.loss_type == 1: #sphere
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
fc7 = mx.sym.LSoftmax(data=embedding,
label=gt_label,
num_hidden=args.num_classes,
weight=_weight,
beta=args.beta,
margin=args.margin,
scale=args.scale,
beta_min=args.beta_min,
verbose=1000,
name='fc7')
elif args.loss_type == 2:
s = args.margin_s
m = args.margin_m
assert (s > 0.0)
assert (m > 0.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
s_m = s * m
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=s_m,
off_value=0.0)
fc7 = fc7 - gt_one_hot
elif args.loss_type == 4:
s = args.margin_s
m = args.margin_m
assert s > 0.0
assert m >= 0.0
assert m < (math.pi / 2)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
cos_m = math.cos(m)
sin_m = math.sin(m)
mm = math.sin(math.pi - m) * m
#threshold = 0.0
threshold = math.cos(math.pi - m)
if args.easy_margin:
cond = mx.symbol.Activation(data=cos_t, act_type='relu')
else:
cond_v = cos_t - threshold
cond = mx.symbol.Activation(data=cond_v, act_type='relu')
body = cos_t * cos_t
body = 1.0 - body
sin_t = mx.sym.sqrt(body)
new_zy = cos_t * cos_m
b = sin_t * sin_m
new_zy = new_zy - b
new_zy = new_zy * s
if args.easy_margin:
zy_keep = zy
else:
zy_keep = zy - s * mm
new_zy = mx.sym.where(cond, new_zy, zy_keep)
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=1.0,
off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
elif args.loss_type == 5:
s = args.margin_s
m = args.margin_m
assert s > 0.0
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
if args.margin_a != 1.0 or args.margin_m != 0.0 or args.margin_b != 0.0:
if args.margin_a == 1.0 and args.margin_m == 0.0:
s_m = s * args.margin_b
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=s_m,
off_value=0.0)
fc7 = fc7 - gt_one_hot
else:
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
t = mx.sym.arccos(cos_t)
if args.margin_a != 1.0:
t = t * args.margin_a
if args.margin_m > 0.0:
t = t + args.margin_m
body = mx.sym.cos(t)
if args.margin_b > 0.0:
body = body - args.margin_b
new_zy = body * s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=1.0,
off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
elif args.loss_type == 6:
s = args.margin_s
m = args.margin_m
assert s > 0.0
assert args.margin_b > 0.0
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
t = mx.sym.arccos(cos_t)
intra_loss = t / np.pi
intra_loss = mx.sym.mean(intra_loss)
#intra_loss = mx.sym.exp(cos_t*-1.0)
intra_loss = mx.sym.MakeLoss(intra_loss,
name='intra_loss',
grad_scale=args.margin_b)
if m > 0.0:
t = t + m
body = mx.sym.cos(t)
new_zy = body * s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=1.0,
off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
elif args.loss_type == 7:
s = args.margin_s
m = args.margin_m
assert s > 0.0
assert args.margin_b > 0.0
assert args.margin_a > 0.0
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
t = mx.sym.arccos(cos_t)
#counter_weight = mx.sym.take(_weight, gt_label, axis=1)
#counter_cos = mx.sym.dot(counter_weight, _weight, transpose_a=True)
counter_weight = mx.sym.take(_weight, gt_label, axis=0)
counter_cos = mx.sym.dot(counter_weight, _weight, transpose_b=True)
#counter_cos = mx.sym.minimum(counter_cos, 1.0)
#counter_angle = mx.sym.arccos(counter_cos)
#counter_angle = counter_angle * -1.0
#counter_angle = counter_angle/np.pi #[0,1]
#inter_loss = mx.sym.exp(counter_angle)
#counter_cos = mx.sym.dot(_weight, _weight, transpose_b=True)
#counter_cos = mx.sym.minimum(counter_cos, 1.0)
#counter_angle = mx.sym.arccos(counter_cos)
#counter_angle = mx.sym.sort(counter_angle, axis=1)
#counter_angle = mx.sym.slice_axis(counter_angle, axis=1, begin=0,end=int(args.margin_a))
#inter_loss = counter_angle*-1.0 # [-1,0]
#inter_loss = inter_loss+1.0 # [0,1]
inter_loss = counter_cos
inter_loss = mx.sym.mean(inter_loss)
inter_loss = mx.sym.MakeLoss(inter_loss,
name='inter_loss',
grad_scale=args.margin_b)
if m > 0.0:
t = t + m
body = mx.sym.cos(t)
new_zy = body * s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=1.0,
off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
out_list = [mx.symbol.BlockGrad(embedding)]
softmax = mx.symbol.SoftmaxOutput(data=fc7,
label=gt_label,
name='softmax',
normalization='valid')
out_list.append(softmax)
if args.loss_type == 6:
out_list.append(intra_loss)
if args.loss_type == 7:
out_list.append(inter_loss)
#out_list.append(mx.sym.BlockGrad(counter_weight))
#out_list.append(intra_loss)
if args.ce_loss:
#ce_loss = mx.symbol.softmax_cross_entropy(data=fc7, label = gt_label, name='ce_loss')/args.per_batch_size
body = mx.symbol.SoftmaxActivation(data=fc7)
body = mx.symbol.log(body)
_label = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=-1.0,
off_value=0.0)
body = body * _label
ce_loss = mx.symbol.sum(body) / args.per_batch_size
out_list.append(mx.symbol.BlockGrad(ce_loss))
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params):
data_shape = (args.image_channel,args.image_h,args.image_w)
image_shape = ",".join([str(x) for x in data_shape])
margin_symbols = []
if args.network[0]=='d':
embedding = fdensenet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='m':
print('init mobilenet', args.num_layers)
if args.num_layers==1:
embedding = fmobilenet.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
elif args.network[0]=='i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0]=='s':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
else:
print('init resnet', args.num_layers)
embedding = fresnet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit,
version_act=args.version_act)
all_label = mx.symbol.Variable('softmax_label')
gt_label = all_label
extra_loss = None
s = args.margin_s
#m = args.margin_m
assert s>0.0
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
out_list = [mx.symbol.BlockGrad(embedding)]
_args = copy.deepcopy(args)
if USE_FR:
_args.grad_scale = 1.0
fr_label = mx.symbol.slice_axis(all_label, axis=1, begin=0, end=1)
fr_label = mx.symbol.reshape(fr_label, (args.per_batch_size,))
fr_softmax = get_softmax(_args, embedding, nembedding, fr_label, 'fc7')
out_list.append(fr_softmax)
if USE_GENDER:
_args.grad_scale = 0.2
_args.margin_a = 0.0
_args.num_classes = 2
gender_label = mx.symbol.slice_axis(all_label, axis=1, begin=1, end=2)
gender_label = mx.symbol.reshape(gender_label, (args.per_batch_size,))
gender_softmax = get_softmax(_args, embedding, nembedding, gender_label, 'fc8')
out_list.append(gender_softmax)
if USE_AGE:
_args.grad_scale = 0.01
_args.margin_a = 0.0
_args.num_classes = 2
for i in xrange(AGE):
age_label = mx.symbol.slice_axis(all_label, axis=1, begin=2+i, end=3+i)
age_label = mx.symbol.reshape(age_label, (args.per_batch_size,))
age_softmax = get_softmax(_args, embedding, nembedding, age_label, 'fc9_%d'%(i))
out_list.append(age_softmax)
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params, sym_embedding=None):
if sym_embedding is None:
if args.network[0] == 'd':
embedding = fdensenet.get_symbol(
args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'm':
print('init mobilenet', args.num_layers)
if args.num_layers == 1:
embedding = fmobilenet.get_symbol(
args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
elif args.network[0] == 'i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(
args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(
args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0] == 's':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
else:
print('init resnet', args.num_layers)
embedding = fresnet.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit,
version_act=args.version_act)
else:
embedding = sym_embedding
gt_label = mx.symbol.Variable('softmax_label')
nembedding = mx.symbol.L2Normalization(embedding,
mode='instance',
name='fc1n')
anchor = mx.symbol.slice_axis(nembedding,
axis=0,
begin=0,
end=args.per_batch_size // 3)
positive = mx.symbol.slice_axis(nembedding,
axis=0,
begin=args.per_batch_size // 3,
end=2 * args.per_batch_size // 3)
negative = mx.symbol.slice_axis(nembedding,
axis=0,
begin=2 * args.per_batch_size // 3,
end=args.per_batch_size)
ap = anchor - positive
an = anchor - negative
ap = ap * ap
an = an * an
ap = mx.symbol.sum(ap, axis=1, keepdims=1) #(T,1)
an = mx.symbol.sum(an, axis=1, keepdims=1) #(T,1)
triplet_loss = mx.symbol.Activation(data=(ap - an + args.triplet_alpha),
act_type='relu')
triplet_loss = mx.symbol.mean(triplet_loss)
#triplet_loss = mx.symbol.sum(triplet_loss)/(args.per_batch_size//3)
triplet_loss = mx.symbol.MakeLoss(triplet_loss)
out_list = [mx.symbol.BlockGrad(embedding)]
out_list.append(mx.sym.BlockGrad(gt_label))
out_list.append(triplet_loss)
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params):
data_shape = (args.image_channel,args.image_h,args.image_w)
image_shape = ",".join([str(x) for x in data_shape])
margin_symbols = []
if args.network[0]=='d':
embedding = fdensenet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='m':
print('init mobilenet', args.num_layers)
if args.num_layers==1:
embedding = fmobilenet.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
elif args.network[0]=='i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0]=='s':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
elif args.network[0]=='y':
print('init mobilefacenet', args.num_layers)
embedding = fmobilefacenet.get_symbol(args.emb_size, bn_mom = args.bn_mom, wd_mult = args.fc7_wd_mult)
else:
print('init resnet', args.num_layers)
embedding = fresnet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit,
version_act=args.version_act)
all_label = mx.symbol.Variable('softmax_label')
gt_label = all_label
extra_loss = None
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0, wd_mult=args.fc7_wd_mult)
if args.loss_type==0: #softmax
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding, weight = _weight, bias = _bias, num_hidden=args.num_classes, name='fc7')
elif args.loss_type==1: #sphere
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
fc7 = mx.sym.LSoftmax(data=embedding, label=gt_label, num_hidden=args.num_classes,
weight = _weight,
beta=args.beta, margin=args.margin, scale=args.scale,
beta_min=args.beta_min, verbose=1000, name='fc7')
elif args.loss_type==2:
s = args.margin_s
m = args.margin_m
assert(s>0.0)
assert(m>0.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
s_m = s*m
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7-gt_one_hot
elif args.loss_type==4:
s = args.margin_s
m = args.margin_m
assert s>0.0
assert m>=0.0
assert m<(math.pi/2)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
cos_m = math.cos(m)
sin_m = math.sin(m)
mm = math.sin(math.pi-m)*m
#threshold = 0.0
threshold = math.cos(math.pi-m)
if args.easy_margin:
cond = mx.symbol.Activation(data=cos_t, act_type='relu')
else:
cond_v = cos_t - threshold
cond = mx.symbol.Activation(data=cond_v, act_type='relu')
body = cos_t*cos_t
body = 1.0-body
sin_t = mx.sym.sqrt(body)
new_zy = cos_t*cos_m
b = sin_t*sin_m
new_zy = new_zy - b
new_zy = new_zy*s
if args.easy_margin:
zy_keep = zy
else:
zy_keep = zy - s*mm
new_zy = mx.sym.where(cond, new_zy, zy_keep)
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7+body
elif args.loss_type==5:
s = args.margin_s
m = args.margin_m
assert s>0.0
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
if args.margin_a!=1.0 or args.margin_m!=0.0 or args.margin_b!=0.0:
if args.margin_a==1.0 and args.margin_m==0.0:
s_m = s*args.margin_b
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7-gt_one_hot
else:
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
t = mx.sym.arccos(cos_t)
if args.margin_a!=1.0:
t = t*args.margin_a
if args.margin_m>0.0:
t = t+args.margin_m
body = mx.sym.cos(t)
if args.margin_b>0.0:
body = body - args.margin_b
new_zy = body*s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7+body
out_list = [mx.symbol.BlockGrad(embedding)]
softmax = mx.symbol.SoftmaxOutput(data=fc7, label = gt_label, name='softmax', normalization='valid')
out_list.append(softmax)
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params):
data_shape = (args.image_channel,args.image_h,args.image_w)
image_shape = ",".join([str(x) for x in data_shape])
margin_symbols = []
if args.network[0]=='d':
embedding = fdensenet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='m':
print('init mobilenet', args.num_layers)
if args.num_layers==1:
embedding = fmobilenet.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
elif args.network[0]=='i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0]=='s':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
elif args.network[0]=='y':
print('init mobilefacenet', args.num_layers)
embedding = fmobilefacenet.get_symbol(args.emb_size)
else:
print('init resnet', args.num_layers)
embedding = fresnet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit,
version_act=args.version_act)
all_label = mx.symbol.Variable('softmax_label')
gt_label = all_label
extra_loss = None
s = args.margin_s
#m = args.margin_m
assert s>0.0
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
out_list = [mx.symbol.BlockGrad(embedding)]
_args = copy.deepcopy(args)
if USE_FR:
_args.grad_scale = 1.0
fr_label = mx.symbol.slice_axis(all_label, axis=1, begin=0, end=1)
fr_label = mx.symbol.reshape(fr_label, (args.per_batch_size,))
fr_softmax = get_softmax(_args, embedding, nembedding, fr_label, 'fc7')
out_list.append(fr_softmax)
if USE_GENDER:
_args.grad_scale = 0.2
_args.margin_a = 0.0
_args.num_classes = 2
gender_label = mx.symbol.slice_axis(all_label, axis=1, begin=1, end=2)
gender_label = mx.symbol.reshape(gender_label, (args.per_batch_size,))
gender_softmax = get_softmax(_args, embedding, nembedding, gender_label, 'fc8')
out_list.append(gender_softmax)
if USE_AGE:
_args.grad_scale = 0.01
_args.margin_a = 0.0
_args.num_classes = 2
for i in xrange(AGE):
age_label = mx.symbol.slice_axis(all_label, axis=1, begin=2+i, end=3+i)
age_label = mx.symbol.reshape(age_label, (args.per_batch_size,))
age_softmax = get_softmax(_args, embedding, nembedding, age_label, 'fc9_%d'%(i))
out_list.append(age_softmax)
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params):
data_shape = (args.image_channel,args.image_h,args.image_w)
image_shape = ",".join([str(x) for x in data_shape])
margin_symbols = []
if args.network[0]=='d':
embedding = fdensenet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='m':
print('init mobilenet', args.num_layers)
if args.num_layers==1:
embedding = fmobilenet.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
elif args.network[0]=='i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0]=='s':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
else:
print('init resnet', args.num_layers)
embedding = fresnet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit,
version_act=args.version_act)
all_label = mx.symbol.Variable('softmax_label')
if not args.output_c2c:
gt_label = all_label
else:
gt_label = mx.symbol.slice_axis(all_label, axis=1, begin=0, end=1)
gt_label = mx.symbol.reshape(gt_label, (args.per_batch_size,))
c2c_label = mx.symbol.slice_axis(all_label, axis=1, begin=1, end=2)
c2c_label = mx.symbol.reshape(c2c_label, (args.per_batch_size,))
assert args.loss_type>=0
extra_loss = None
if args.loss_type==0: #softmax
_weight = mx.symbol.Variable('fc7_weight')
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding, weight = _weight, bias = _bias, num_hidden=args.num_classes, name='fc7')
elif args.loss_type==1: #sphere
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
fc7 = mx.sym.LSoftmax(data=embedding, label=gt_label, num_hidden=args.num_classes,
weight = _weight,
beta=args.beta, margin=args.margin, scale=args.scale,
beta_min=args.beta_min, verbose=1000, name='fc7')
elif args.loss_type==8: #centerloss, TODO
_weight = mx.symbol.Variable('fc7_weight')
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding, weight = _weight, bias = _bias, num_hidden=args.num_classes, name='fc7')
print('center-loss', args.center_alpha, args.center_scale)
extra_loss = mx.symbol.Custom(data=embedding, label=gt_label, name='center_loss', op_type='centerloss',\
num_class=args.num_classes, alpha=args.center_alpha, scale=args.center_scale, batchsize=args.per_batch_size)
elif args.loss_type==2:
s = args.margin_s
m = args.margin_m
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
if s>0.0:
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
if m>0.0:
if args.margin_verbose>0:
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
margin_symbols.append(mx.symbol.mean(cos_t))
s_m = s*m
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7-gt_one_hot
if args.margin_verbose>0:
new_zy = mx.sym.pick(fc7, gt_label, axis=1)
new_cos_t = new_zy/s
margin_symbols.append(mx.symbol.mean(new_cos_t))
else:
fc7 = mx.sym.FullyConnected(data=embedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
if m>0.0:
body = embedding*embedding
body = mx.sym.sum_axis(body, axis=1, keepdims=True)
body = mx.sym.sqrt(body)
body = body*m
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, body)
fc7 = fc7-body
elif args.loss_type==3:
s = args.margin_s
m = args.margin_m
assert args.margin==2 or args.margin==4
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
if args.margin_verbose>0:
margin_symbols.append(mx.symbol.mean(cos_t))
if m>1.0:
t = mx.sym.arccos(cos_t)
t = t*m
body = mx.sym.cos(t)
new_zy = body*s
if args.margin_verbose>0:
new_cos_t = new_zy/s
margin_symbols.append(mx.symbol.mean(new_cos_t))
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7+body
#threshold = math.cos(args.margin_m)
#cond_v = cos_t - threshold
#cond = mx.symbol.Activation(data=cond_v, act_type='relu')
#body = cos_t
#for i in xrange(args.margin//2):
# body = body*body
# body = body*2-1
#new_zy = body*s
#zy_keep = zy
#new_zy = mx.sym.where(cond, new_zy, zy_keep)
#if args.margin_verbose>0:
# new_cos_t = new_zy/s
# margin_symbols.append(mx.symbol.mean(new_cos_t))
#diff = new_zy - zy
#diff = mx.sym.expand_dims(diff, 1)
#gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
#body = mx.sym.broadcast_mul(gt_one_hot, diff)
#fc7 = fc7+body
elif args.loss_type==4:
s = args.margin_s
m = args.margin_m
assert s>0.0
assert m>=0.0
assert m<(math.pi/2)
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
if args.margin_verbose>0:
margin_symbols.append(mx.symbol.mean(cos_t))
if args.output_c2c==0:
cos_m = math.cos(m)
sin_m = math.sin(m)
mm = math.sin(math.pi-m)*m
#threshold = 0.0
threshold = math.cos(math.pi-m)
if args.easy_margin:
cond = mx.symbol.Activation(data=cos_t, act_type='relu')
else:
cond_v = cos_t - threshold
cond = mx.symbol.Activation(data=cond_v, act_type='relu')
body = cos_t*cos_t
body = 1.0-body
sin_t = mx.sym.sqrt(body)
new_zy = cos_t*cos_m
b = sin_t*sin_m
new_zy = new_zy - b
new_zy = new_zy*s
if args.easy_margin:
zy_keep = zy
else:
zy_keep = zy - s*mm
new_zy = mx.sym.where(cond, new_zy, zy_keep)
else:
#set c2c as cosm^2 in data.py
cos_m = mx.sym.sqrt(c2c_label)
sin_m = 1.0-c2c_label
sin_m = mx.sym.sqrt(sin_m)
body = cos_t*cos_t
body = 1.0-body
sin_t = mx.sym.sqrt(body)
new_zy = cos_t*cos_m
b = sin_t*sin_m
new_zy = new_zy - b
new_zy = new_zy*s
if args.margin_verbose>0:
new_cos_t = new_zy/s
margin_symbols.append(mx.symbol.mean(new_cos_t))
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7+body
elif args.loss_type==5:
#s = args.margin_s
#m = args.margin_m
#assert s>0.0
#assert m>=0.0
#assert m<(math.pi/2)
#_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
#_weight = mx.symbol.L2Normalization(_weight, mode='instance')
#nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
#fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
#zy = mx.sym.pick(fc7, gt_label, axis=1)
#cos_t = zy/s
#if args.margin_verbose>0:
# margin_symbols.append(mx.symbol.mean(cos_t))
#if m>0.0:
# a1 = args.margin_a
# r1 = ta-a1
# r1 = mx.symbol.Activation(data=r1, act_type='relu')
# r1 = r1+a1
# t = mx.sym.arccos(cos_t)
# cond = t-1.0
# cond = mx.symbol.Activation(data=cond, act_type='relu')
# r = mx.sym.where(cond, r2, r1)
# t = t+var_m
# body = mx.sym.cos(t)
# new_zy = body*s
# if args.margin_verbose>0:
# new_cos_t = new_zy/s
# margin_symbols.append(mx.symbol.mean(new_cos_t))
# #margin_symbols.append(mx.symbol.mean(var_m))
# diff = new_zy - zy
# diff = mx.sym.expand_dims(diff, 1)
# gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
# body = mx.sym.broadcast_mul(gt_one_hot, diff)
# fc7 = fc7+body
s = args.margin_s
m = args.margin_m
assert s>0.0
#assert m>=0.0
#assert m<(math.pi/2)
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
t = mx.sym.arccos(cos_t)
if args.margin_verbose>0:
margin_symbols.append(mx.symbol.mean(t))
if args.margin_a>0.0:
t = t*args.margin_a
if args.margin_m>0.0:
t = t+args.margin_m
body = mx.sym.cos(t)
if args.margin_b>0.0:
body = body - args.margin_b
new_zy = body*s
if args.margin_verbose>0:
margin_symbols.append(mx.symbol.mean(t))
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7+body
elif args.loss_type==6:
s = args.margin_s
m = args.margin_m
assert s>0.0
assert m>=0.0
assert m<(math.pi/2)
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
t = mx.sym.arccos(cos_t)
if args.margin_verbose>0:
margin_symbols.append(mx.symbol.mean(t))
t_min = mx.sym.min(t)
ta = mx.sym.broadcast_div(t_min, t)
a1 = args.margin_a
r1 = ta-a1
r1 = mx.symbol.Activation(data=r1, act_type='relu')
r1 = r1+a1
r2 = mx.symbol.zeros(shape=(args.per_batch_size,))
cond = t-1.0
cond = mx.symbol.Activation(data=cond, act_type='relu')
r = mx.sym.where(cond, r2, r1)
var_m = r*m
t = t+var_m
body = mx.sym.cos(t)
new_zy = body*s
if args.margin_verbose>0:
#new_cos_t = new_zy/s
#margin_symbols.append(mx.symbol.mean(new_cos_t))
margin_symbols.append(mx.symbol.mean(t))
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7+body
elif args.loss_type==7:
s = args.margin_s
m = args.margin_m
assert s>0.0
assert m>=0.0
assert m<(math.pi/2)
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
t = mx.sym.arccos(cos_t)
if args.margin_verbose>0:
margin_symbols.append(mx.symbol.mean(t))
var_m = mx.sym.random.uniform(low=args.margin_a, high=args.margin_m, shape=(1,))
t = mx.sym.broadcast_add(t,var_m)
body = mx.sym.cos(t)
new_zy = body*s
if args.margin_verbose>0:
#new_cos_t = new_zy/s
#margin_symbols.append(mx.symbol.mean(new_cos_t))
margin_symbols.append(mx.symbol.mean(t))
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7+body
elif args.loss_type==10: #marginal loss
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')
params = [1.2, 0.3, 1.0]
n1 = mx.sym.expand_dims(nembedding, axis=1) #N,1,C
n2 = mx.sym.expand_dims(nembedding, axis=0) #1,N,C
body = mx.sym.broadcast_sub(n1, n2) #N,N,C
body = body * body
body = mx.sym.sum(body, axis=2) # N,N
#body = mx.sym.sqrt(body)
body = body - params[0]
mask = mx.sym.Variable('extra')
body = body*mask
body = body+params[1]
#body = mx.sym.maximum(body, 0.0)
body = mx.symbol.Activation(data=body, act_type='relu')
body = mx.sym.sum(body)
body = body/(args.per_batch_size*args.per_batch_size-args.per_batch_size)
extra_loss = mx.symbol.MakeLoss(body, grad_scale=params[2])
elif args.loss_type==11: #npair loss
params = [0.9, 0.2]
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')
nembedding = mx.sym.transpose(nembedding)
nembedding = mx.symbol.reshape(nembedding, (args.emb_size, args.per_identities, args.images_per_identity))
nembedding = mx.sym.transpose(nembedding, axes=(2,1,0)) #2*id*512
#nembedding = mx.symbol.reshape(nembedding, (args.emb_size, args.images_per_identity, args.per_identities))
#nembedding = mx.sym.transpose(nembedding, axes=(1,2,0)) #2*id*512
n1 = mx.symbol.slice_axis(nembedding, axis=0, begin=0, end=1)
n2 = mx.symbol.slice_axis(nembedding, axis=0, begin=1, end=2)
#n1 = []
#n2 = []
#for i in xrange(args.per_identities):
# _n1 = mx.symbol.slice_axis(nembedding, axis=0, begin=2*i, end=2*i+1)
# _n2 = mx.symbol.slice_axis(nembedding, axis=0, begin=2*i+1, end=2*i+2)
# n1.append(_n1)
# n2.append(_n2)
#n1 = mx.sym.concat(*n1, dim=0)
#n2 = mx.sym.concat(*n2, dim=0)
#rembeddings = mx.symbol.reshape(nembedding, (args.images_per_identity, args.per_identities, 512))
#n1 = mx.symbol.slice_axis(rembeddings, axis=0, begin=0, end=1)
#n2 = mx.symbol.slice_axis(rembeddings, axis=0, begin=1, end=2)
n1 = mx.symbol.reshape(n1, (args.per_identities, args.emb_size))
n2 = mx.symbol.reshape(n2, (args.per_identities, args.emb_size))
cosine_matrix = mx.symbol.dot(lhs=n1, rhs=n2, transpose_b = True) #id*id, id=N of N-pair
data_extra = mx.sym.Variable('extra')
data_extra = mx.sym.slice_axis(data_extra, axis=0, begin=0, end=args.per_identities)
mask = cosine_matrix * data_extra
#body = mx.sym.mean(mask)
fii = mx.sym.sum_axis(mask, axis=1)
fij_fii = mx.sym.broadcast_sub(cosine_matrix, fii)
fij_fii = mx.sym.exp(fij_fii)
row = mx.sym.sum_axis(fij_fii, axis=1)
row = mx.sym.log(row)
body = mx.sym.mean(row)
extra_loss = mx.sym.MakeLoss(body)
elif args.loss_type==12: #triplet loss
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')
anchor = mx.symbol.slice_axis(nembedding, axis=0, begin=0, end=args.per_batch_size//3)
positive = mx.symbol.slice_axis(nembedding, axis=0, begin=args.per_batch_size//3, end=2*args.per_batch_size//3)
negative = mx.symbol.slice_axis(nembedding, axis=0, begin=2*args.per_batch_size//3, end=args.per_batch_size)
ap = anchor - positive
an = anchor - negative
ap = ap*ap
an = an*an
ap = mx.symbol.sum(ap, axis=1, keepdims=1) #(T,1)
an = mx.symbol.sum(an, axis=1, keepdims=1) #(T,1)
triplet_loss = mx.symbol.Activation(data = (ap-an+args.triplet_alpha), act_type='relu')
triplet_loss = mx.symbol.mean(triplet_loss)
#triplet_loss = mx.symbol.sum(triplet_loss)/(args.per_batch_size//3)
extra_loss = mx.symbol.MakeLoss(triplet_loss)
elif args.loss_type==13: #triplet loss with angular margin
m = args.margin_m
sin_m = math.sin(m)
cos_m = math.cos(m)
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')
anchor = mx.symbol.slice_axis(nembedding, axis=0, begin=0, end=args.per_batch_size//3)
positive = mx.symbol.slice_axis(nembedding, axis=0, begin=args.per_batch_size//3, end=2*args.per_batch_size//3)
negative = mx.symbol.slice_axis(nembedding, axis=0, begin=2*args.per_batch_size//3, end=args.per_batch_size)
ap = anchor * positive
an = anchor * negative
ap = mx.symbol.sum(ap, axis=1, keepdims=1) #(T,1)
an = mx.symbol.sum(an, axis=1, keepdims=1) #(T,1)
ap = mx.symbol.arccos(ap)
an = mx.symbol.arccos(an)
triplet_loss = mx.symbol.Activation(data = (ap-an+args.margin_m), act_type='relu')
#body = ap*ap
#body = 1.0-body
#body = mx.symbol.sqrt(body)
#body = body*sin_m
#ap = ap*cos_m
#ap = ap-body
#triplet_loss = mx.symbol.Activation(data = (an-ap), act_type='relu')
triplet_loss = mx.symbol.mean(triplet_loss)
extra_loss = mx.symbol.MakeLoss(triplet_loss)
elif args.loss_type==9: #coco loss
centroids = []
for i in xrange(args.per_identities):
xs = mx.symbol.slice_axis(embedding, axis=0, begin=i*args.images_per_identity, end=(i+1)*args.images_per_identity)
mean = mx.symbol.mean(xs, axis=0, keepdims=True)
mean = mx.symbol.L2Normalization(mean, mode='instance')
centroids.append(mean)
centroids = mx.symbol.concat(*centroids, dim=0)
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*args.coco_scale
fc7 = mx.symbol.dot(nembedding, centroids, transpose_b = True) #(batchsize, per_identities)
#extra_loss = mx.symbol.softmax_cross_entropy(fc7, gt_label, name='softmax_ce')/args.per_batch_size
#extra_loss = mx.symbol.BlockGrad(extra_loss)
else:
#embedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*float(args.loss_type)
embedding = embedding * 5
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance') * 2
fc7 = mx.sym.LSoftmax(data=embedding, label=gt_label, num_hidden=args.num_classes,
weight = _weight,
beta=args.beta, margin=args.margin, scale=args.scale,
beta_min=args.beta_min, verbose=100, name='fc7')
#fc7 = mx.sym.Custom(data=embedding, label=gt_label, weight=_weight, num_hidden=args.num_classes,
# beta=args.beta, margin=args.margin, scale=args.scale,
# op_type='ASoftmax', name='fc7')
if args.loss_type<=1 and args.incay>0.0:
params = [1.e-10]
sel = mx.symbol.argmax(data = fc7, axis=1)
sel = (sel==gt_label)
norm = embedding*embedding
norm = mx.symbol.sum(norm, axis=1)
norm = norm+params[0]
feature_incay = sel/norm
feature_incay = mx.symbol.mean(feature_incay) * args.incay
extra_loss = mx.symbol.MakeLoss(feature_incay)
#out = softmax
#l2_embedding = mx.symbol.L2Normalization(embedding)
#ce = mx.symbol.softmax_cross_entropy(fc7, gt_label, name='softmax_ce')/args.per_batch_size
#out = mx.symbol.Group([mx.symbol.BlockGrad(embedding), softmax, mx.symbol.BlockGrad(ce)])
out_list = [mx.symbol.BlockGrad(embedding)]
softmax = None
if args.loss_type<10:
softmax = mx.symbol.SoftmaxOutput(data=fc7, label = gt_label, name='softmax', normalization='valid')
out_list.append(softmax)
if args.logits_verbose>0:
logits = mx.symbol.softmax(data = fc7)
logits = mx.sym.pick(logits, gt_label, axis=1)
margin_symbols.append(logits)
#logit_max = mx.sym.max(logits)
#logit_min = mx.sym.min(logits)
#margin_symbols.append(logit_max)
#margin_symbols.append(logit_min)
if softmax is None:
out_list.append(mx.sym.BlockGrad(gt_label))
if extra_loss is not None:
out_list.append(extra_loss)
for _sym in margin_symbols:
_sym = mx.sym.BlockGrad(_sym)
out_list.append(_sym)
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params):
# data_shape = (args.image_channel, args.image_h, args.image_w) # (3L,112L,112L)
# image_shape = ",".join([str(x) for x in data_shape]) #3,112,112
# margin_symbols = []
print('***network: ', args.network) # r100
if args.network[0] == 'd': # densenet
embedding = fdensenet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0] == 'm': # mobilenet
print('init mobilenet', args.num_layers)
if args.num_layers == 1:
embedding = fmobilenet.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
# elif args.network[0] == 'v':
# print('init MobileNet-V3', args.num_layers)
# embedding = fmobilenetv3.get_symbol(args.emb_size)
elif args.network[0] == 'i': # inception-resnet-v2
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0] == 'x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0] == 'p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0] == 'n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0] == 's':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
elif args.network[0] == 'y':
print('init mobilefacenet', args.num_layers)
embedding = fmobilefacenet.get_symbol(args.emb_size, bn_mom=args.bn_mom, version_output=args.version_output)
else: # 执行resnet
print('init resnet, 层数: ', args.num_layers)
embedding = fresnet.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit,
version_act=args.version_act)
# get_symbol
all_label = mx.symbol.Variable('softmax_label')
gt_label = all_label
# extra_loss = None
# 重新定义fc7的权重
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=args.fc7_lr_mult,
wd_mult=args.fc7_wd_mult)
if args.loss_type == 0: # softmax
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding, weight=_weight, bias=_bias, num_hidden=args.num_classes, name='fc7')
elif args.loss_type == 1: # sphere
print('*******'*10)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
fc7 = mx.sym.LSoftmax(data=embedding, label=gt_label, num_hidden=args.num_classes,
weight=_weight,
beta=args.beta, margin=args.margin, scale=args.scale,
beta_min=args.beta_min, verbose=1000, name='fc7')
elif args.loss_type == 2: # CosineFace
s = args.margin_s
m = args.margin_m
assert (s > 0.0)
assert (m > 0.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding, weight=_weight, no_bias=True, num_hidden=args.num_classes,
name='fc7')
s_m = s * m
gt_one_hot = mx.sym.one_hot(gt_label, depth=args.num_classes, on_value=s_m,
off_value=0.0) # onehot两个值最大值s_m,最小值0.0
fc7 = fc7 - gt_one_hot
elif args.loss_type == 4: # ArcFace
s = args.margin_s # 参数s， 64
m = args.margin_m # 参数m， 0.5
assert s > 0.0
assert m >= 0.0
assert m < (math.pi / 2)
# pdb.set_trace()
# 权重归一化
_weight = mx.symbol.L2Normalization(_weight, mode='instance') # shape = [(4253, 512)]
# 特征归一化，并放大到 s*x
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding, weight=_weight, no_bias=True, num_hidden=args.num_classes,
name='fc7') # args.num_classes:8631
zy = mx.sym.pick(fc7, gt_label, axis=1) # fc7每一行找出gt_label对应的值, 即s*cos_t
cos_t = zy / s # 网络输出output = s*x/|x|*w/|w|*cos(theta), 这里将输出除以s，得到实际的cos值，即cos（theta)
cos_m = math.cos(m)
sin_m = math.sin(m)
mm = math.sin(math.pi - m) * m # sin(pi-m)*m = sin(m) * m 0.2397
# threshold = 0.0
threshold = math.cos(math.pi - m) # 这个阈值避免theta+m >= pi, 实际上threshold<0 -cos(m) -0.8775825618903726
if args.easy_margin: # 将0作为阈值，得到超过阈值的索引
cond = mx.symbol.Activation(data=cos_t, act_type='relu')
else:
cond_v = cos_t - threshold # 将负数作为阈值
cond = mx.symbol.Activation(data=cond_v, act_type='relu')
body = cos_t * cos_t # cos_t^2 + sin_t^2 = 1
body = 1.0 - body
sin_t = mx.sym.sqrt(body)
new_zy = cos_t * cos_m # cos(t+m) = cos(t)cos(m) - sin(t)sin(m)
b = sin_t * sin_m
new_zy = new_zy - b
new_zy = new_zy * s # s*cos(t + m)
if args.easy_margin:
zy_keep = zy # zy_keep为zy，即s*cos(theta)
else:
zy_keep = zy - s * mm # zy-s*sin(m)*m = s*cos(t)- s*m*sin(m)
new_zy = mx.sym.where(cond, new_zy,
zy_keep) # cond中>0的保持new_zy=s*cos(theta+m)不变，<0的裁剪为zy_keep= s*cos(theta) or s*cos(theta)-s*m*sin(m)
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth=args.num_classes, on_value=1.0, off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff) # 对应yi处为new_zy - zy
fc7 = fc7 + body # 对应yi处，fc7=zy + (new_zy - zy) = new_zy，即cond中>0的为s*cos(theta+m)，<0的裁剪为s*cos(theta) or s*cos(theta)-s*m*sin(m)
elif args.loss_type == 5:
s = args.margin_s
m = args.margin_m
assert s > 0.0
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding, weight=_weight, no_bias=True, num_hidden=args.num_classes,
name='fc7')
if args.margin_a != 1.0 or args.margin_m != 0.0 or args.margin_b != 0.0:
if args.margin_a == 1.0 and args.margin_m == 0.0:
s_m = s * args.margin_b
gt_one_hot = mx.sym.one_hot(gt_label, depth=args.num_classes, on_value=s_m, off_value=0.0)
fc7 = fc7 - gt_one_hot
else:
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
t = mx.sym.arccos(cos_t)
if args.margin_a != 1.0:
t = t * args.margin_a
if args.margin_m > 0.0:
t = t + args.margin_m
body = mx.sym.cos(t)
if args.margin_b > 0.0:
body = body - args.margin_b
new_zy = body * s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth=args.num_classes, on_value=1.0, off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
out_list = [mx.symbol.BlockGrad(embedding)]
softmax = mx.symbol.SoftmaxOutput(data=fc7, label=gt_label, name='softmax', normalization='valid')
out_list.append(softmax)
out = mx.symbol.Group(out_list)
# print(out)
# sys.exit()
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params):
data_shape = (args.image_channel, args.image_h, args.image_w)
image_shape = ",".join([str(x) for x in data_shape])
margin_symbols = []
if args.network[0] == 'd':
embedding = fdensenet.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'm':
print('init mobilenet', args.num_layers)
if args.num_layers == 1:
embedding = fmobilenet.get_symbol(
args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
elif args.network[0] == 'i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(
args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(args.emb_size,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit)
elif args.network[0] == 'n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0] == 's':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
elif args.network[0] == 'y':
print('init mobilefacenet', args.num_layers)
embedding = fmobilefacenet.get_symbol(args.emb_size,
bn_mom=args.bn_mom,
wd_mult=args.fc7_wd_mult)
else:
print('init resnet', args.num_layers)
embedding = fresnet.get_symbol(args.emb_size,
args.num_layers,
version_se=args.version_se,
version_input=args.version_input,
version_output=args.version_output,
version_unit=args.version_unit,
version_act=args.version_act)
all_label = mx.symbol.Variable('softmax_label')
gt_label = all_label
extra_loss = None
_weight = mx.symbol.Variable("fc7_weight",
shape=(args.num_classes, args.emb_size),
lr_mult=1.0,
wd_mult=args.fc7_wd_mult)
if args.loss_type == 0: #softmax
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding,
weight=_weight,
bias=_bias,
num_hidden=args.num_classes,
name='fc7')
elif args.loss_type == 1: #sphere
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
fc7 = mx.sym.LSoftmax(data=embedding,
label=gt_label,
num_hidden=args.num_classes,
weight=_weight,
beta=args.beta,
margin=args.margin,
scale=args.scale,
beta_min=args.beta_min,
verbose=1000,
name='fc7')
elif args.loss_type == 2:
s = args.margin_s
m = args.margin_m
assert (s > 0.0)
assert (m > 0.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
s_m = s * m
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=s_m,
off_value=0.0)
fc7 = fc7 - gt_one_hot
elif args.loss_type == 4:
s = args.margin_s
m = args.margin_m
assert s > 0.0
assert m >= 0.0
assert m < (math.pi / 2)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
# split fc7 into fc7(labeled part) and fc7_2(unlabeled part)
# calculate the likelihood function of the fc7_2 part
subtract = args.num_classes * math.log(
args.num_classes
) ############### This is calculated based on the number of labeled classes
num_unlabeled = args.num_unlabeled
num_labeled = args.per_batch_size - args.num_unlabeled
per_batch_size = args.per_batch_size
assert num_labeled > 0
fc7_2 = mx.sym.slice_axis(fc7,
axis=0,
begin=num_labeled,
end=per_batch_size)
fc7 = mx.sym.slice_axis(fc7, axis=0, begin=0, end=num_labeled)
fc7_2 = mx.sym.softmax(data=fc7_2, axis=1)
log_likelihood = -mx.sym.log_softmax(fc7_2, axis=1)
log_likelihood_loss = (mx.sym.sum(log_likelihood) / num_unlabeled)
log_likelihood_loss = log_likelihood_loss - subtract
gt_label = mx.sym.slice_axis(gt_label,
axis=0,
begin=0,
end=num_labeled)
############################################################
#last_to_append = mx.symbol.BlockGrad(mx.sym.softmax(fc7))
out_list = []
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
cos_m = math.cos(m)
sin_m = math.sin(m)
mm = math.sin(math.pi - m) * m
#threshold = 0.0
threshold = math.cos(math.pi - m)
if args.easy_margin:
cond = mx.symbol.Activation(data=cos_t, act_type='relu')
else:
cond_v = cos_t - threshold
cond = mx.symbol.Activation(data=cond_v, act_type='relu')
body = cos_t * cos_t
body = 1.0 - body
sin_t = mx.sym.sqrt(body)
new_zy = cos_t * cos_m
b = sin_t * sin_m
new_zy = new_zy - b
new_zy = new_zy * s
if args.easy_margin:
zy_keep = zy
else:
zy_keep = zy - s * mm
new_zy = mx.sym.where(cond, new_zy, zy_keep)
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=1.0,
off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
elif args.loss_type == 5:
s = args.margin_s
m = args.margin_m
assert s > 0.0
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(
embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(data=nembedding,
weight=_weight,
no_bias=True,
num_hidden=args.num_classes,
name='fc7')
if args.margin_a != 1.0 or args.margin_m != 0.0 or args.margin_b != 0.0:
if args.margin_a == 1.0 and args.margin_m == 0.0:
s_m = s * args.margin_b
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=s_m,
off_value=0.0)
fc7 = fc7 - gt_one_hot
else:
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
t = mx.sym.arccos(cos_t)
if args.margin_a != 1.0:
t = t * args.margin_a
if args.margin_m > 0.0:
t = t + args.margin_m
body = mx.sym.cos(t)
if args.margin_b > 0.0:
body = body - args.margin_b
new_zy = body * s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label,
depth=args.num_classes,
on_value=1.0,
off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
out_list.append(mx.symbol.BlockGrad(embedding))
if args.loss_type == 4:
#softmax = mx.sym.softmax_cross_entropy(data=fc7, label=gt_label)
cos_fc7 = mx.symbol.BlockGrad(
fc7) # TODO: maybe this softmax makes the accuracy inconsistent
gt_label_one_hot = mx.sym.one_hot(gt_label, args.num_classes)
fc7 = -mx.sym.log_softmax(fc7, axis=1)
softmax = gt_label_one_hot * fc7
softmax = mx.sym.sum(softmax) / (args.per_batch_size -
args.num_unlabeled)
scale = args.likelihood_scale_factor
loss = mx.sym.make_loss(data=(softmax + log_likelihood_loss * scale))
out_list.append(loss)
out_list.append(cos_fc7)
out_list.append(mx.sym.BlockGrad(softmax))
out_list.append(mx.sym.BlockGrad(log_likelihood_loss))
else:
softmax = mx.symbol.SoftmaxOutput(data=fc7,
label=gt_label,
name='softmax',
normalization='valid')
out_list.append(softmax)
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params):
data_shape = (args.image_channel,args.image_h,args.image_w)
image_shape = ",".join([str(x) for x in data_shape])
margin_symbols = []
if args.network[0]=='d':
embedding = fdensenet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='m':
print('init mobilenet', args.num_layers)
if args.num_layers==1:
embedding = fmobilenet.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
elif args.network[0]=='i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0]=='s':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
elif args.network[0]=='y':
print('init mobilefacenet', args.num_layers)
embedding = fmobilefacenet.get_symbol(args.emb_size, bn_mom = args.bn_mom, wd_mult = args.fc7_wd_mult)
else:
print('init resnet', args.num_layers)
embedding = fresnet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit,
version_act=args.version_act)
all_label = mx.symbol.Variable('softmax_label')
#center_label = mx.symbol.Variable('center_label')
gt_label = all_label
extra_loss = None
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0, wd_mult=args.fc7_wd_mult)
if args.loss_type==0: #softmax
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(data=embedding, weight = _weight, bias = _bias, num_hidden=args.num_classes, name='fc7')
elif args.loss_type==1: #sphere
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
fc7 = mx.sym.LSoftmax(data=embedding, label=gt_label, num_hidden=args.num_classes,
weight = _weight,
beta=args.beta, margin=args.margin, scale=args.scale,
beta_min=args.beta_min, verbose=1000, name='fc7')
elif args.loss_type==2:
s = args.margin_s
m = args.margin_m
assert(s>0.0)
assert(m>0.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
s_m = s*m
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7-gt_one_hot
elif args.loss_type==4:
s = args.margin_s
m = args.margin_m
assert s>0.0
assert m>=0.0
assert m<(math.pi/2)
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
t = mx.sym.arccos(cos_t)
if args.margin_a!=1.0:
t = t*args.margin_a
if args.margin_m>0.0:
t = t+args.margin_m
body = mx.sym.cos(t)
if args.margin_b>0.0:
body = body - args.margin_b
new_zy = body*s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7+body
elif args.loss_type==5:
s = args.margin_s
m = args.margin_m
assert s>0.0
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
nembedding = mx.symbol.Dropout(data=nembedding, p=0.4)
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
if args.margin_a!=1.0 or args.margin_m!=0.0 or args.margin_b!=0.0:
if args.margin_a==1.0 and args.margin_m==0.0:
s_m = s*args.margin_b
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7-gt_one_hot
else:
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
t = mx.sym.arccos(cos_t)
if args.margin_a!=1.0:
t = t*args.margin_a
if args.margin_m>0.0:
t = t+args.margin_m
#threshold = math.cos(math.pi-m)
if args.easy_margin: # m<pi/2 so: pi-m > pi/2
cond = mx.symbol.Activation(data=cos_t, act_type='relu') # this means t > pi/2, when pi/2 <t < 3*pi/2, sin is decrease
#cond_v = cos_t - threshold
#cond = mx.symbol.Activation(data=cond_v, act_type='relu')
else:
cond_v = math.pi - t
cond = mx.symbol.Activation(data=cond_v, act_type='relu')
print("not easy margin: cond")
#cond = mx.symbol.Activation(data=cos_t, act_type='relu')
if args.easy_margin:
zy_keep = mx.sym.cos(t)
else:
print("not easy margin: sine")
zy_keep = mx.sym.sin(t)-1
new_zy = mx.sym.cos(t)
body = mx.sym.where(cond, new_zy, zy_keep)
#body = mx.sym.cos(t)
if args.margin_b>0.0:
body = body + args.margin_b
new_zy = body*s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7+body
elif args.loss_type==6:
s = args.margin_s
m = args.margin_m
assert s>0.0
_weight = mx.symbol.L2Normalization(_weight, mode='instance')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(data=nembedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
if args.margin_a!=1.0 or args.margin_m!=0.0 or args.margin_b!=0.0:
if args.margin_a==1.0 and args.margin_m==0.0:
s_m = s*args.margin_b
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7-gt_one_hot
else:
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
t = mx.sym.arccos(cos_t)
if args.margin_a!=1.0:
t = t*args.margin_a
if args.margin_m>0.0:
t = t+args.margin_m
body = mx.sym.cos(t)
if args.margin_b>0.0:
body = body - args.margin_b
new_zy = body*s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7+body
out_list = [mx.symbol.BlockGrad(embedding)]
softmax = mx.symbol.SoftmaxOutput(data=fc7, label = gt_label, name='softmax', normalization='valid')
out_list.append(softmax)
#print("out shape",np.shape(softmax))
#center_in = mx.symbol.concat(embedding,fc7,dim=1)
#center_loss_data = mx.symbol.Custom(data=embedding, label=gt_label, name='center_loss_data', op_type='centerloss',\
# num_class=args.num_classes, alpha=0.5, scale=0.5,lamdb=0.1,batchsize=args.per_batch_size,emb_size=args.emb_size)
#extra_center_loss = mx.symbol.MakeLoss(name='extra_center_loss', data=center_loss_data)
#total_loss = mx.symbol.ElementWiseSum([softmax, extra_loss],name='total_loss')
#total_loss_op = mx.symbol.MakeLoss(name='total_loss_op',data=total_loss)
#out_list.append(extra_center_loss)
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)
def get_symbol(args, arg_params, aux_params):
data_shape = (args.image_channel,args.image_h,args.image_w)
image_shape = ",".join([str(x) for x in data_shape])
margin_symbols = []
if args.network[0]=='d':
embedding = fdensenet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='m':
print('init mobilenet', args.num_layers)
if args.num_layers==1:
embedding = fmobilenet.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
else:
embedding = fmobilenetv2.get_symbol(args.emb_size)
elif args.network[0]=='i':
print('init inception-resnet-v2', args.num_layers)
embedding = finception_resnet_v2.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='x':
print('init xception', args.num_layers)
embedding = fxception.get_symbol(args.emb_size,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='p':
print('init dpn', args.num_layers)
embedding = fdpn.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit)
elif args.network[0]=='n':
print('init nasnet', args.num_layers)
embedding = fnasnet.get_symbol(args.emb_size)
elif args.network[0]=='s':
print('init spherenet', args.num_layers)
embedding = spherenet.get_symbol(args.emb_size, args.num_layers)
elif args.network[0] == 'c':
print('init crunet', args.num_layers)
embedding = fcrunet.get_symbol(args.emb_size, args.num_layers)
elif args.network[0] == 'a':
print('init residual attention network', args.num_layers)
embedding = fresattnet.get_symbol(args.emb_size, args.num_layers)
else:
print('init resnet', args.num_layers)
ibn = args.ibn
if ibn:
embedding = fresnet_ibn_a.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit,
version_act=args.version_act, ibn=args.ibn)
else:
embedding = fresnet.get_symbol(args.emb_size, args.num_layers,
version_se=args.version_se, version_input=args.version_input,
version_output=args.version_output, version_unit=args.version_unit,
version_act=args.version_act, stn=args.stn, stn1=args.stn1, stn2=args.stn2, stn3=args.stn3, stn4=args.stn4)
all_label = mx.symbol.Variable('softmax_label')
gt_label = all_label
extra_loss = None
if args.loss_type==0: #softmax
_weight = mx.symbol.Variable('fc7_weight')
_bias = mx.symbol.Variable('fc7_bias', lr_mult=2.0, wd_mult=0.0)
fc7 = mx.sym.FullyConnected(
data=embedding, weight=_weight,
bias=_bias, num_hidden=args.num_classes,
name='fc7'
)
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
elif args.loss_type==1: #sphereface
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(
_weight, mode='instance', name='fc7_weight_n')
fc7 = mx.sym.LSoftmax(
data=embedding, label=gt_label,
num_hidden=args.num_classes, weight = _weight,
beta=args.beta, margin=args.margin,
scale=args.scale, beta_min=args.beta_min,
verbose=1000, name='fc7_lsoftmax'
)
# for softmax ACC computation
fc_pred = mx.sym.FullyConnected(
data=embedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7'
)
elif args.loss_type==2:
s = args.margin_s
m = args.margin_m
assert(s>0.0)
assert(m>0.0)
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(
_weight, mode='instance', name='fc7_weight_n')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7'
)
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
s_m = s*m
gt_one_hot = mx.sym.one_hot(
gt_label, depth=args.num_classes,
on_value=s_m, off_value=0.0)
fc7 = fc7 - gt_one_hot
elif args.loss_type==4: #ArcFace
s = args.margin_s
m = args.margin_m
assert s>0.0
assert m>=0.0
assert m < (math.pi / 2)
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight, no_bias=True,
num_hidden=args.num_classes, name='fc7'
)
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
cos_m = math.cos(m)
sin_m = math.sin(m)
mm = math.sin(math.pi-m)*m
#threshold = 0.0
threshold = math.cos(math.pi-m)
if args.easy_margin:
cond = mx.symbol.Activation(data=cos_t, act_type='relu')
else:
cond_v = cos_t - threshold
cond = mx.symbol.Activation(data=cond_v, act_type='relu')
body = cos_t*cos_t
body = 1.0-body
sin_t = mx.sym.sqrt(body)
new_zy = cos_t*cos_m
b = sin_t*sin_m
new_zy = new_zy - b
new_zy = new_zy * s
if args.easy_margin:
zy_keep = zy
else:
zy_keep = zy - s*mm
new_zy = mx.sym.where(cond, new_zy, zy_keep)
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
elif args.loss_type==5:
s = args.margin_s
m = args.margin_m
assert s>0.0
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7')
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
if args.margin_a!=1.0 or args.margin_m!=0.0 or args.margin_b!=0.0:
if args.margin_a==1.0 and args.margin_m==0.0:
s_m = s*args.margin_b
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7-gt_one_hot
else:
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
t = mx.sym.arccos(cos_t)
if args.margin_a!=1.0:
t = t*args.margin_a
if args.margin_m>0.0:
t = t+args.margin_m
body = mx.sym.cos(t)
if args.margin_b>0.0:
body = body - args.margin_b
new_zy = body*s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7+body
elif args.loss_type == 6: #spa loss
s = args.margin_s
m = args.margin_m
b = args.margin_b
assert (s >= 1.0 and
m >= 1.0 and
b >= 0.0)
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
if s>0.0:
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7')
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
if m>1.0:
s_m = s*(m - 1 + b )
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7 * m - gt_one_hot
else:
fc7 = mx.sym.FullyConnected(data=embedding, weight = _weight, no_bias = True, num_hidden=args.num_classes, name='fc7')
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
if m>1.0:
body = embedding*embedding
body = mx.sym.sum_axis(body, axis=1, keepdims=True)
body = mx.sym.sqrt(body)
body = body * (m - 1 + b)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1.0, off_value = 0.0)
body = mx.sym.broadcast_mul(gt_one_hot, body)
fc7 = fc7 * m - body
elif args.loss_type == 61: #spa-v1 loss fixed
s = args.margin_s
m = args.margin_m
b = args.margin_b
assert (s >= 1.0 and
m >= 1.0 and
b >= 0.0)
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7')
fc_pred = fc7
if m > 1.0:
cos_theta = fc7.clip(-s, s) / s # clip cosine into [-1, 1]
cos_theta_1 = cos_theta - 1
# s_m = s*(m - 1 + b )
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = 1, off_value = 0.0)
cos_theta = cos_theta_1 + cos_theta_1 * gt_one_hot * (m - 1)
if b > 0:
cos_theta -= gt_one_hot * b
fc7 = cos_theta * s
elif args.loss_type == 63: # spa-v3 loss
s = args.margin_s
m = args.margin_m
assert (s >= 1.0 and
m >= 0.0)
_weight = mx.symbol.Variable("fc7_weight", shape=(
args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7')
fc_pred = fc7
if m > 0:
cos_theta = fc7.clip(-s, s) / s # clip cosine into [-1, 1]
theta = mx.sym.arccos(cos_theta)
gt_one_hot = mx.sym.one_hot(
gt_label, depth=args.num_classes, on_value=1, off_value=0.0)
cos_theta = cos_theta - gt_one_hot * theta * m
fc7 = cos_theta * s
elif args.loss_type == 64: # spa-v4 loss
s = args.margin_s
m = args.margin_m
b = args.margin_b
if b < 1.0:
b=1.0
assert (s >= 1.0 and
m >= b and
b >= 1.0)
_weight = mx.symbol.Variable("fc7_weight", shape=(
args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7')
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
cos_theta = fc7.clip(-s, s) / s # clip cosine into [-1, 1]
# calculate angle theta and normalize into [0, 1.0]
theta = mx.sym.arccos(cos_theta) * (1.0 / np.pi)
s *= 2 # make the output have a region span of 2 like cos(x)
if m > b:
gt_one_hot = mx.sym.one_hot(
gt_label, depth=args.num_classes, on_value=1, off_value=0.0)
fc7 = (theta * b + gt_one_hot * theta * (m-b)) * (-s)
else:
fc7 = theta * (-s)
elif args.loss_type == 65: # spa-v5 loss
s = args.margin_s
m = args.margin_m
assert (s >= 1.0 and
m >= 0 and
m < 180)
_weight = mx.symbol.Variable("fc7_weight", shape=(
args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
m = m / 180.0
nembedding = mx.symbol.L2Normalization( embedding, mode='instance', name='fc1n')
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight, no_bias=True, num_hidden=args.num_classes, name='fc7')
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
cos_theta = fc7.clip(-s, s) / s # clip cosine into [-1, 1]
# calculate angle theta and normalize into [0, 1.0]
theta = mx.sym.arccos(cos_theta) * (1.0 / np.pi)
s *= 2 # make the output have a region span of 2 like cos(x)
if m > 0:
gt_one_hot = mx.sym.one_hot(
gt_label, depth=args.num_classes, on_value=1, off_value=0.0)
fc7 = (theta + gt_one_hot * m) * (-s)
else:
fc7 = theta * (-s)
elif args.loss_type == 7: #combine spa loss
s = args.margin_s
m = args.margin_m
b = args.margin_b
a = args.margin_a
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
assert m>=1.0
assert b>=0.0
assert s>0.0
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7')
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
t = mx.sym.arccos(cos_t)
t = t * a
new_zy = mx.sym.cos(t) * s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth=args.num_classes, on_value=1.0, off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
s_m = s*(m - 1)
gt_one_hot2 = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = m * (fc7 + body) - gt_one_hot2
elif args.loss_type == 8: #Adaloss
s = args.margin_s
m = args.margin_m
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n') * s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7')
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy / s
t = mx.sym.arccos(cos_t)
body = ((1-mx.sym.sin(1.0*t/(2*m)))* mx.sym.cos(1.0 *t/m) * 2) - 1
new_zy = body * s
diff = new_zy - zy
diff = mx.sym.expand_dims(diff, 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth=args.num_classes, on_value=1.0, off_value=0.0)
body = mx.sym.broadcast_mul(gt_one_hot, diff)
fc7 = fc7 + body
elif args.loss_type == 9: #semi hard loss, have fault
s = args.margin_s
m = args.margin_m
b = args.margin_b
a = args.margin_a
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
assert m>=1.0
assert b>=0.0
if s>0.0:
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7')
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
t = mx.sym.arccos(cos_t)
bounding = mx.sym.Variable('')
s_m = s*(m - 1 + b )
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7 * m - gt_one_hot
elif args.loss_type==10:#combine intra loss
s = args.margin_s
m = args.margin_m
assert s>0.0
assert args.margin_b>0.0
b = args.margin_b
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7')
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=nembedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
zy = mx.sym.pick(fc7, gt_label, axis=1)
cos_t = zy/s
t = mx.sym.arccos(cos_t)
intra_loss = t/np.pi
intra_loss = mx.sym.mean(intra_loss)
intra_loss = mx.sym.MakeLoss(intra_loss, name='intra_loss', grad_scale = args.margin_b)
s_m = s*(m - 1)
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7 * m - gt_one_hot
elif args.loss_type == 11: #reweight
s = args.margin_s
m = args.margin_m
b = args.margin_b
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
assert m>=1.0
assert b>=0.0
#reweight
spatial_norm = embedding * embedding
spatial_norm = mx.sym.sum(data=spatial_norm, axis=1, keepdims=True)
spatial_sqrt = mx.sym.sqrt(spatial_norm)
spatial_mean = mx.sym.mean(spatial_sqrt)
spatial_div_inverse = mx.sym.broadcast_div(spatial_mean, spatial_sqrt)
reweight_s = s * spatial_div_inverse
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')
# for softmax ACC computation
fc_pred = mx.sym.FullyConnected(
data=embedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc_pred'
) * s
# fc_pred = fc7
nembedding = mx.symbol.broadcast_mul(nembedding, reweight_s)
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7')
s_m = s*(m - 1 + b )
gt_one_hot = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7 * m - gt_one_hot
elif args.loss_type == 12: #hard example margin
s = args.margin_s
m = args.margin_m
b = args.margin_b
_weight = mx.symbol.Variable("fc7_weight", shape=(args.num_classes, args.emb_size), lr_mult=1.0)
_weight = mx.symbol.L2Normalization(_weight, mode='instance', name='fc7_weight_n')
assert m>=1.0
assert b>=0.0
nembedding = mx.symbol.L2Normalization(embedding, mode='instance', name='fc1n')*s
fc7 = mx.sym.FullyConnected(
data=nembedding, weight=_weight,
no_bias=True, num_hidden=args.num_classes,
name='fc7')
# for softmax ACC computation
# fc_pred = mx.sym.FullyConnected(
# data=embedding, weight=_weight,
# no_bias=True, num_hidden=args.num_classes,
# name='fc_pred'
# )
fc_pred = fc7
predict_label = mx.sym.argmax(fc7, axis=1)
wrong_label_mask = predict_label.__lt__(gt_label)
wrong_label = predict_label * wrong_label_mask
s_m = s*(m - 1 + b )
wrong_label_one_hot = mx.sym.one_hot(wrong_label, depth = args.num_classes, on_value = s_m, off_value = 0.0)
fc7 = fc7 * m - wrong_label_one_hot
out_list = [mx.symbol.BlockGrad(embedding)]
softmax = mx.symbol.SoftmaxOutput(
data=fc7, label=gt_label,
name='softmax', normalization='valid')
out_list.append(softmax)
# for softmax ACC computation
pred_softmax = mx.symbol.SoftmaxOutput(
data=fc_pred, label=gt_label,
name='pred_softmax', normalization='valid')
out_list.append(mx.symbol.BlockGrad(pred_softmax))
out_list.append(mx.symbol.BlockGrad(_weight))
if args.loss_type == 10:
out_list.append(intra_loss)
if args.ce_loss:
#ce_loss = mx.symbol.softmax_cross_entropy(data=fc7, label = gt_label, name='ce_loss')/args.per_batch_size
body = mx.symbol.SoftmaxActivation(data=fc7)
body = mx.symbol.log(body)
_label = mx.sym.one_hot(gt_label, depth = args.num_classes, on_value = -1.0, off_value = 0.0)
body = body*_label
ce_loss = mx.symbol.sum(body)/args.per_batch_size
out_list.append(mx.symbol.BlockGrad(ce_loss))
out = mx.symbol.Group(out_list)
return (out, arg_params, aux_params)