def get_embeddings(image, net, device):
transform = cvtransforms.Compose([
cvtransforms.Resize((112, 112)),
cvtransforms.RandomHorizontalFlip(),
cvtransforms.ToTensor(),
cvtransforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))
])
transformed_image = transform(image).to(device)
the_image = Variable(transformed_image).unsqueeze(0)
# net.eval()
embeddings = l2_norm(net.forward(the_image)).detach() # remain data at gpu
return embeddings
test_indices, trainer):
pass
def parameters_dict_to_model_name(parameters_dict):
pass
if __name__ == "__main__":
img_path = args.img_path
gt_path = args.gt_path
# create dataset
train_input_transform_list = [
cvtransforms.Resize(size=input_tensor_res, interpolation='BILINEAR'),
cvtransforms.RandomHorizontalFlip(),
cvtransforms.RandomVerticalFlip(),
cvtransforms.RandomRotation(90),
cvtransforms.ToTensor(),
# cvtransforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]
val_input_transform_list = [
cvtransforms.Resize(size=input_tensor_res, interpolation='BILINEAR'),
cvtransforms.ToTensor(),
# cvtransforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
]
train_transforms = [
compose_input_output_transform(
input_transform=cvtransforms.Compose(train_input_transform_list)),