def res50_train(mean_value, list_file, is_train=True):
# setup the python data layer
net = caffe.NetSpec()
net.data, net.label \
= L.ReidData(transform_param=dict(mirror=True,crop_size=224,mean_value=mean_value),
reid_data_param=dict(source=list_file,batch_size=16,new_height=256,new_width=256,
pos_fraction=1,neg_fraction=1,pos_limit=1,neg_limit=4,pos_factor=1,neg_factor=1.01),
ntop = 2)
net, final = res50_body(net, 'data', '', is_train)
net['score'] = fc_relu(net[final],
nout=751,
is_train=is_train,
has_relu=False)
net['euclidean'], net['label_dif'] = L.PairEuclidean(net[final],
net['label'],
ntop=2)
net['score_dif'] = fc_relu(net['euclidean'],
nout=2,
is_train=is_train,
has_relu=False)
net['loss'] = L.SoftmaxWithLoss(net['score'],
net['label'],
propagate_down=[1, 0],
loss_weight=0.5)
net['loss_dif'] = L.SoftmaxWithLoss(net['score_dif'],
net['label_dif'],
propagate_down=[1, 0],
loss_weight=1)
return str(net.to_proto())
def get_reidDataLayer(net, batch_size=1):
kwargs = {
# 'include': dict(phase=caffe_pb2.Phase.Value('TRAIN')),
'reid_transform_param': get_reidTransParam(),
}
reid_data_kwargs = get_reidDataParam(batch_size=batch_size)
net.data, net.label = L.ReidData(name="data", reid_data_param=reid_data_kwargs, \
ntop=2, **kwargs)
return net
def res50_train(mean_value, list_file, is_train, batch_size):
# setup the python data layer
net = caffe.NetSpec()
net.data, net.label \
= L.ReidData(transform_param=dict(mirror=True,crop_size=224,mean_value=mean_value),
reid_data_param=dict(source=list_file,batch_size=batch_size, new_height=256, new_width=256,
pos_fraction=1,neg_fraction=1,pos_limit=1,neg_limit=4,pos_factor=1, neg_factor=1.01),
ntop = 2)
net, final = res50_body(net, 'data', '', is_train)
param = [dict(lr_mult=1, decay_mult=1), dict(lr_mult=2, decay_mult=0)]
net['score'] = fc_relu(net[final],
nout=751,
is_train=is_train,
has_relu=False,
param=param)
#net['euclidean'], net['label_dif'] = L.PairEuclidean(net[final], net['label'], ntop = 2)
net['label_dif'] = L.PairReidLabel(net['label'],
propagate_down=[0],
ntop=1)
net['feature_a'], net['feature_b'] = L.Slice(net[final],
slice_param=dict(
axis=0,
slice_point=batch_size),
ntop=2)
net['euclidean'] = L.Eltwise(net['feature_a'],
net['feature_b'],
operation=P.Eltwise.PROD)
net['score_dif'] = fc_relu(net['euclidean'],
nout=2,
is_train=is_train,
has_relu=False,
param=param)
net['loss'] = L.SoftmaxWithLoss(net['score'],
net['label'],
propagate_down=[1, 0],
loss_weight=0.5)
net['loss_dif'] = L.SoftmaxWithLoss(net['score_dif'],
net['label_dif'],
propagate_down=[1, 0],
loss_weight=1)
return str(net.to_proto())
def CreateDataLayer_SV_IMG(source,
batch_size=28,
new_width=224,
new_height=224,
pos_fraction=1,
neg_fraction=1,
pos_limit=1.0,
neg_limit=4.0,
pos_factor=1.0,
neg_factor=1.01,
is_color=True,
train=True,
transform_param={}):
if train:
kwargs = {
'include': dict(phase=caffe_pb2.Phase.Value('TRAIN')),
'transform_param': transform_param,
}
else:
kwargs = {
'include': dict(phase=caffe_pb2.Phase.Value('TEST')),
'transform_param': transform_param,
}
return L.ReidData(name="data",
reid_data_param=dict(source=source,
new_width=new_width,
new_height=new_height,
pos_fraction=1,
neg_fraction=1,
pos_limit=1.0,
neg_limit=4.0,
pos_factor=1.0,
neg_factor=1.01,
is_color=is_color,
batch_size=batch_size),
ntop=2,
**kwargs)
