def simple_train_preprocessed(wdecay,
lr,
route_iter,
folds=[4, 9],
model_name='vgg_capsule_disfa',
epoch_stuff=[30, 60],
res=False):
out_dirs = []
out_dir_meta = '../experiments/ck_khcap_recon_r' + str(
route_iter) + '_noinit'
num_epochs = epoch_stuff[1]
epoch_start = 0
dec_after = ['step', epoch_stuff[0], 0.1]
# data/Oulu_CASIA/train_test_files_preprocess_vl/
lr = lr
im_resize = 110
im_size = 96
# model_name = 'vgg_capsule_disfa'
save_after = 10
# type_data = 'three_im_no_neutral_just_strong_False'; n_classes = 6;
type_data = 'train_test_files'
n_classes = 8
# ../data/Oulu_CASIA/train_test_files_preprocess_maheen_vl_gray/
strs_append = '_' + '_'.join([
str(val)
for val in [model_name, 'all_aug', 'wdecay', wdecay, num_epochs] +
dec_after + lr
])
pre_pend = 'ck_' + type_data + '_'
for split_num in folds:
if res:
strs_appendc = '_'.join([
str(val) for val in
[model_name, 'all_aug', 'wdecay', wdecay, 50, 'step', 50, 0.1
] + lr
])
out_dir_train = os.path.join(
out_dir_meta, 'oulu_' + type_data + '_' + str(split_num) +
'_' + strs_appendc)
model_file = os.path.join(out_dir_train, 'model_49.pt')
epoch_start = 50
else:
model_file = None
criterion = 'margin'
margin_params = None
out_dir_train = os.path.join(out_dir_meta,
pre_pend + str(split_num) + strs_append)
final_model_file = os.path.join(out_dir_train,
'model_' + str(num_epochs - 1) + '.pt')
if os.path.exists(final_model_file):
print 'skipping', final_model_file
continue
# train_file = os.path.join('../data/Oulu_CASIA','train_test_files_preprocess_maheen_vl_gray',type_data,'train_'+str(split_num)+'.txt')
# test_file = os.path.join('../data/Oulu_CASIA','train_test_files_preprocess_maheen_vl_gray',type_data,'test_'+str(split_num)+'.txt')
train_file = os.path.join('../data/ck_96', type_data,
'train_' + str(split_num) + '.txt')
test_file = os.path.join('../data/ck_96', type_data,
'test_' + str(split_num) + '.txt')
mean_std = np.array([[128.], [1.]]) #bgr
print mean_std
data_transforms = {}
data_transforms['train'] = transforms.Compose([
transforms.ToPILImage(),
transforms.Resize((im_resize, im_resize)),
transforms.RandomCrop(im_size),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(15),
transforms.ColorJitter(),
transforms.ToTensor(),
lambda x: x * 255,
transforms.Normalize(mean_std[0], mean_std[1]),
])
data_transforms['val'] = transforms.Compose([
transforms.ToPILImage(),
# transforms.Resize((im_size,im_size)),
transforms.ToTensor(),
lambda x: x * 255,
transforms.Normalize(mean_std[0, :], mean_std[1, :]),
])
train_data = dataset.Oulu_Static_Dataset(train_file,
data_transforms['train'],
color=False)
test_data = dataset.Oulu_Static_Dataset(test_file,
data_transforms['val'],
color=False)
network_params = dict(n_classes=n_classes,
loss=nn.CrossEntropyLoss(),
in_size=im_size,
r=route_iter,
init=False)
# if lr[0]==0:
batch_size = 128
batch_size_val = 128
# else:
# batch_size = 32
# batch_size_val = 16
util.makedirs(out_dir_train)
train_params = dict(out_dir_train=out_dir_train,
train_data=train_data,
test_data=test_data,
batch_size=batch_size,
batch_size_val=batch_size_val,
num_epochs=num_epochs,
save_after=save_after,
disp_after=1,
plot_after=10,
test_after=1,
lr=lr,
dec_after=dec_after,
model_name=model_name,
criterion=criterion,
gpu_id=0,
num_workers=0,
model_file=model_file,
epoch_start=epoch_start,
margin_params=margin_params,
network_params=network_params,
weight_decay=wdecay)
print train_params
param_file = os.path.join(out_dir_train, 'params.txt')
all_lines = []
for k in train_params.keys():
str_print = '%s: %s' % (k, train_params[k])
print str_print
all_lines.append(str_print)
util.writeFile(param_file, all_lines)
train_model(**train_params)
test_params = dict(out_dir_train=out_dir_train,
model_num=num_epochs - 1,
train_data=train_data,
test_data=test_data,
gpu_id=0,
model_name=model_name,
batch_size_val=batch_size_val,
criterion=criterion,
margin_params=margin_params,
network_params=network_params)
test_model(**test_params)
getting_accuracy.print_accuracy(out_dir_meta,
pre_pend,
strs_append,
folds,
log='log.txt')
def simple_train_preprocessed(wdecay,
lr,
route_iter,
folds=[4, 9],
model_name='vgg_capsule_disfa',
epoch_stuff=[30, 60],
res=False):
out_dir_meta = '../experiments/oulu_vgg_r' + str(
route_iter) + '_noinit_preprocessed'
num_epochs = epoch_stuff[1]
epoch_start = 0
dec_after = ['step', epoch_stuff[0], 0.1]
# data/Oulu_CASIA/train_test_files_preprocess_vl/
lr = lr
im_resize = 256
im_size = 224
# model_name = 'vgg_capsule_disfa'
save_after = 10
type_data = 'three_im_no_neutral_just_strong_False'
n_classes = 6
strs_append = '_' + '_'.join([
str(val)
for val in [model_name, 'all_aug', 'wdecay', wdecay, num_epochs] +
dec_after + lr
])
pre_pend = 'oulu_' + type_data + '_'
for split_num in folds:
if res:
strs_append_c = '_'.join([
str(val) for val in
[model_name, 'all_aug', 'wdecay', wdecay, 50, 'step', 50, 0.1
] + lr
])
out_dir_train = os.path.join(
out_dir_meta, 'oulu_' + type_data + '_' + str(split_num) +
'_' + strs_append_c)
model_file = os.path.join(out_dir_train, 'model_49.pt')
epoch_start = 50
else:
model_file = None
# model_file = '../experiments/oulu_vgg_r3_noinit/oulu_three_im_no_neutral_just_strong_False_4_vgg_capsule_disfa_bigprimary_all_aug_wdecay_0_100_step_100_0.1_0.0001_0.001/model_30.pt'
# epoch_start = 30
# type_data = 'single_im'
criterion = torch.nn.CrossEntropyLoss()
margin_params = None
# strs_append = '_'.join([str(val) for val in [model_name,'all_aug','wdecay',wdecay,num_epochs]+dec_after+lr])
out_dir_train = os.path.join(out_dir_meta,
pre_pend + str(split_num) + strs_append)
# print out_dir_train
final_model_file = os.path.join(out_dir_train,
'model_' + str(num_epochs - 1) + '.pt')
if os.path.exists(final_model_file):
# print 'skipping',final_model_file
continue
# lr[1]=lr[1]*dec_after[2]
train_file = os.path.join('../data/Oulu_CASIA',
'train_test_files_preprocess_vl', type_data,
'train_' + str(split_num) + '.txt')
test_file = os.path.join('../data/Oulu_CASIA',
'train_test_files_preprocess_vl', type_data,
'test_' + str(split_num) + '.txt')
# mean_std = np.array([[129.1863,104.7624,93.5940],[1.,1.,1.]]) #rgb
mean_std = np.array([[128., 128., 128.], [1., 1., 1.]]) #bgr
# print mean_std
data_transforms = {}
data_transforms['train'] = transforms.Compose([
transforms.ToPILImage(),
transforms.Resize((im_resize, im_resize)),
transforms.RandomCrop(im_size),
transforms.RandomHorizontalFlip(),
transforms.RandomRotation(15),
transforms.ColorJitter(),
transforms.ToTensor(),
lambda x: x * 255,
transforms.Normalize(mean_std[0, :], mean_std[1, :]),
# lambda x: x[[2, 1, 0],:, : ]
])
data_transforms['val'] = transforms.Compose([
transforms.ToPILImage(),
transforms.Resize((im_size, im_size)),
transforms.ToTensor(),
lambda x: x * 255,
transforms.Normalize(mean_std[0, :], mean_std[1, :]),
# lambda x: x[[2, 1, 0],:, : ]
])
data_transforms['val_center'] = transforms.Compose([
transforms.ToPILImage(),
# transforms.Resize((im_resize,im_resize)),
transforms.CenterCrop(im_size),
transforms.ToTensor(),
lambda x: x * 255,
transforms.Normalize(mean_std[0, :], mean_std[1, :]),
# lambda x: x[[2, 1, 0],:, : ]
])
train_data = dataset.Oulu_Static_Dataset(train_file,
data_transforms['train'],
color=True)
test_data = dataset.Oulu_Static_Dataset(test_file,
data_transforms['val'],
color=True)
test_data_center = dataset.Oulu_Static_Dataset(
test_file, data_transforms['val_center'], color=True)
network_params = dict(n_classes=n_classes, r=route_iter, init=False)
if lr[0] == 0:
batch_size = 128
batch_size_val = 128
else:
batch_size = 32
batch_size_val = 16
util.makedirs(out_dir_train)
train_params = dict(out_dir_train=out_dir_train,
train_data=train_data,
test_data=test_data,
batch_size=batch_size,
batch_size_val=batch_size_val,
num_epochs=num_epochs,
save_after=save_after,
disp_after=1,
plot_after=10,
test_after=1,
lr=lr,
dec_after=dec_after,
model_name=model_name,
criterion=criterion,
gpu_id=0,
num_workers=0,
model_file=model_file,
epoch_start=epoch_start,
margin_params=margin_params,
network_params=network_params,
weight_decay=wdecay)
print train_params
param_file = os.path.join(out_dir_train, 'params.txt')
all_lines = []
for k in train_params.keys():
str_print = '%s: %s' % (k, train_params[k])
print str_print
all_lines.append(str_print)
util.writeFile(param_file, all_lines)
train_model(**train_params)
test_params = dict(out_dir_train=out_dir_train,
model_num=num_epochs - 1,
train_data=train_data,
test_data=test_data,
gpu_id=0,
model_name=model_name,
batch_size_val=batch_size_val,
criterion=criterion,
margin_params=margin_params,
network_params=network_params)
test_model(**test_params)
test_params['test_data'] = test_data_center
test_params['post_pend'] = '_center'
test_model(**test_params)
getting_accuracy.print_accuracy(out_dir_meta,
pre_pend,
strs_append,
folds,
log='log.txt')
def trying_out_recon(wdecay, lr):
for split_num in [4, 9]:
out_dir_meta = '../experiments/oulu_with_recon_0.5_r3/'
route_iter = 3
num_epochs = 100
epoch_start = 0
# dec_after = ['exp',0.96,3,1e-6]
dec_after = ['step', 100, 0.1]
lr = lr
# [0.001]
pool_type = 'max'
im_size = 96
model_name = 'khorrami_capsule_reconstruct'
save_after = 50
type_data = 'three_im_no_neutral_just_strong'
n_classes = 6
# strs_append = '_'.join([str(val) for val in ['all_aug','wdecay',wdecay,pool_type,500,'step',500,0.1]+lr])
# out_dir_train = os.path.join(out_dir_meta,'oulu_'+type_data+'_'+str(split_num)+'_'+strs_append)
# model_file = os.path.join(out_dir_train,'model_499.pt')
# type_data = 'single_im'
model_file = None
criterion = 'margin'
margin_params = None
spread_loss_params = dict(end_epoch=int(num_epochs * 0.5),
decay_steps=5,
init_margin=0.5,
max_margin=0.5)
# spread_loss_params = {'end_epoch':int(num_epochs*0.9),'decay_steps':5,'init_margin' : 0.9, 'max_margin' : 0.9}
strs_append = '_'.join([
str(val)
for val in ['justflip', 'wdecay', wdecay, pool_type, num_epochs] +
dec_after + lr
])
out_dir_train = os.path.join(
out_dir_meta,
'oulu_' + type_data + '_' + str(split_num) + '_' + strs_append)
print out_dir_train
train_file = os.path.join(
'../data/Oulu_CASIA', 'train_test_files_preprocess_maheen_vl_gray',
type_data, 'train_' + str(split_num) + '.txt')
test_file = os.path.join('../data/Oulu_CASIA',
'train_test_files_preprocess_maheen_vl_gray',
type_data, 'test_' + str(split_num) + '.txt')
mean_std_file = os.path.join(
'../data/Oulu_CASIA', 'train_test_files_preprocess_maheen_vl_gray',
type_data, 'train_' + str(split_num) + '_mean_std_val_0_1.npy')
class_weights = util.get_class_weights(
util.readLinesFromFile(train_file))
mean_std = np.load(mean_std_file)
print mean_std
data_transforms = {}
data_transforms['train'] = transforms.Compose([
transforms.ToPILImage(),
transforms.Resize((im_size, im_size)),
# transforms.Resize((102,102)),
# transforms.RandomCrop(im_size),
transforms.RandomHorizontalFlip(),
# transforms.RandomRotation(15),
# transforms.ColorJitter(),
transforms.ToTensor(),
transforms.Normalize([float(mean_std[0])], [float(mean_std[1])])
])
data_transforms['val'] = transforms.Compose([
transforms.ToPILImage(),
transforms.Resize((im_size, im_size)),
transforms.ToTensor(),
transforms.Normalize([float(mean_std[0])], [float(mean_std[1])])
])
# data_transforms['val_center']= transforms.Compose([
# transforms.ToPILImage(),
# transforms.Resize((102,102)),
# transforms.CenterCrop(im_size),
# transforms.ToTensor(),
# transforms.Normalize([float(mean_std[0])],[float(mean_std[1])])
# ])
train_data = dataset.Oulu_Static_Dataset(train_file,
data_transforms['train'])
test_data = dataset.Oulu_Static_Dataset(test_file,
data_transforms['val'])
# test_data_center = dataset.Oulu_Static_Dataset(test_file, data_transforms['val_center'])
network_params = dict(n_classes=n_classes,
spread_loss_params=spread_loss_params,
pool_type=pool_type,
r=route_iter,
init=False,
reconstruct=True,
class_weights=class_weights)
batch_size = 64
batch_size_val = 64
util.makedirs(out_dir_train)
train_params = dict(out_dir_train=out_dir_train,
train_data=train_data,
test_data=test_data,
batch_size=batch_size,
batch_size_val=batch_size_val,
num_epochs=num_epochs,
save_after=save_after,
disp_after=1,
plot_after=10,
test_after=1,
lr=lr,
dec_after=dec_after,
model_name=model_name,
criterion=criterion,
gpu_id=0,
num_workers=0,
model_file=model_file,
epoch_start=epoch_start,
margin_params=margin_params,
network_params=network_params,
weight_decay=wdecay)
print train_params
param_file = os.path.join(out_dir_train, 'params.txt')
all_lines = []
for k in train_params.keys():
str_print = '%s: %s' % (k, train_params[k])
print str_print
all_lines.append(str_print)
util.writeFile(param_file, all_lines)
train_model(**train_params)
test_params = dict(out_dir_train=out_dir_train,
model_num=num_epochs - 1,
train_data=train_data,
test_data=test_data,
gpu_id=0,
model_name=model_name,
batch_size_val=batch_size_val,
criterion=criterion,
margin_params=margin_params,
network_params=network_params)
test_model(**test_params)