def __getitem__(self, index):
# this is what ImageFolder normally returns
original_tuple = super(ImageFolderWithAges, self).__getitem__(index)
age = path2age(self.imgs[index][0], self.pat, self.pos)
# make a new tuple that includes original and the path
tuple_with_path = (original_tuple + (age, ))
return tuple_with_path
def __getitem__(self, index):
# this is what ImageFolder normally returns
original_tuple = super(ImageFolderWithAgeGroup, self).__getitem__(index)
img = self.transforms_(image=np.array(original_tuple[0]))["image"]
age = path2age(self.imgs[index][0], self.pat, self.pos)
# make a new tuple that includes original and the path
tuple_with_path = (img, original_tuple[1], self.find_group(age))
return tuple_with_path
def main():
dataset = mixture
model = DAL_model('cosface', dataset['n_cls'])
if len(sys.argv) >= 4:
model.load_state_dict(torch.load(sys.argv[3]))
print(f'Loaded weights: {sys.argv[3]}')
start_epoch = path2age(sys.argv[3], '_|\.', 0) + 1
else:
start_epoch = 0
trainer = Trainer(model, dataset, int(sys.argv[1]), print_freq=1)
save = '/data/fuzhuolin/DAL/state_dicts/1'
trainer.train(int(sys.argv[2]), start_epoch, save)
def main():
dataset = mixture
model = DAL_model('cosface', dataset['n_cls'])
if len(sys.argv) >= 4:
model.load_state_dict(torch.load(sys.argv[3]))
print(f'Loaded weights: {sys.argv[3]}')
start_epoch = path2age(sys.argv[3], '_|\.', 0) + 1
else:
start_epoch = 0
trainer = Trainer(model, dataset, int(sys.argv[1]), print_freq=1)
save = 'E:/2021WIN/SI681/Decorrelated-Adversarial-Learning/state_dicts/1'
trainer.train(int(sys.argv[2]), start_epoch, save)
def main():
dataset = cacd
model = DAL_model('cosface', dataset['n_cls'])
model.load_state_dict(torch.load("../cache/dal-pretrained-1st-stage.pth"))
# return
if len(sys.argv) >= 4:
model.load_state_dict(torch.load(sys.argv[3]))
print(f'Loaded weights: {sys.argv[3]}')
start_epoch = path2age(sys.argv[3], '_|\.', 0) + 1
else:
start_epoch = 0
trainer = Trainer(model,
dataset,
int(sys.argv[1]),
print_freq=100,
train_head_only=False)
save = '../cache/model_cache/'
trainer.train(int(sys.argv[2]), start_epoch, save)