fasterRCNN = resnet(range(81),
18,
pretrained=False,
class_agnostic=args.class_agnostic)
else:
fasterRCNN = resnet(pascal_classes,
18,
pretrained=False,
class_agnostic=args.class_agnostic)
elif args.net == 'squeeze':
fasterRCNN = squeeze(pascal_classes,
pretrained=True,
class_agnostic=args.class_agnostic)
elif args.net == 'alex':
fasterRCNN = alex(pascal_classes,
pretrained=True,
class_agnostic=args.class_agnostic)
else:
print("network is not defined")
pdb.set_trace()
fasterRCNN.create_architecture()
# print(fasterRCNN)
print("load checkpoint %s" % (load_name))
if args.cuda > 0:
checkpoint = torch.load(load_name)
else:
checkpoint = torch.load(load_name,
map_location=(lambda storage, loc: storage))
fasterRCNN.load_state_dict(checkpoint['model'])
fasterRCNN = resnet(imdb.classes,
10,
pretrained=True,
class_agnostic=args.class_agnostic)
elif args.net == 'res152':
fasterRCNN = resnet(imdb.classes,
152,
pretrained=True,
class_agnostic=args.class_agnostic)
elif args.net == 'squeeze':
fasterRCNN = squeeze(imdb.classes,
pretrained=True,
class_agnostic=args.class_agnostic)
elif args.net == 'alex':
fasterRCNN = alex(imdb.classes,
pretrained=True,
class_agnostic=args.class_agnostic)
else:
print("network is not defined")
pdb.set_trace()
fasterRCNN.create_architecture()
lr = cfg.TRAIN.LEARNING_RATE
lr = args.lr
params = []
for key, value in dict(fasterRCNN.named_parameters()).items():
if value.requires_grad:
if 'bias' in key:
params += [{'params':[value],'lr':lr*(cfg.TRAIN.DOUBLE_BIAS + 1), \