LinMu7177
commented
4 months ago I have removed the condition checking for cur_code == 0 in the decode_sean() method. Is this appropriate? I would appreciate any advice!
def decode_sean(sean_model, image_code, target_mask): obj_dic = load_average_feature()
for idx in range(19):
cur_code = image_code[0, idx]
**_# if not torch.all(cur_code == 0):
# obj_dic[str(idx)]['ACE'] = cur_code
obj_dic[str(idx)]['ACE'] = cur_code_**
temp_face_image = torch.zeros((0, 3, 256, 256)) # place holder
data = {'label': target_mask,
'instance': torch.tensor(0),
'image': temp_face_image.clone().detach(),
'obj_dic': obj_dic}
change_status(sean_model, 'UI_mode')
generated = sean_model(data, mode='UI_mode')[0] # [3, 256, 256]
return generated
maximkm
First of all, thank you for open-sourcing such an outstanding project. When I tried to run the blending_gen.py script to generate data, I encountered the following error：
/home/wenwen/anaconda3/envs/hair/bin/python /home/wenwen/workspace/HairFastGAN/scripts/blending_gen.py Loading StyleGAN2 from checkpoint: ../pretrained_models/StyleGAN/ffhq.pt Loading e4e over the pSp framework from checkpoint: ../pretrained_models/encoder4editing/e4e_ffhq_encode.pt Network [SPADEGenerator] was created. Total number of parameters: 266.9 million. To see the architecture, do print(network). 0it [00:01, ?it/s] Traceback (most recent call last): File "/home/wenwen/workspace/HairFastGAN/scripts/blending_gen.py", line 80, in
main(args)
File "/home/wenwen/workspace/HairFastGAN/scripts/blending_gen.py", line 64, in main
align_shape, align_color, name_to_embed = hair_fast(pt1, pt2, pt3, align_flag=True)
File "/home/wenwen/workspace/HairFastGAN/hair_swap.py", line 105, in call
return self.swap(*args, kwargs)
File "/home/wenwen/workspace/HairFastGAN/hair_swap.py", line 97, in swap
final_image = self.__swap_from_tensors(*images, seed=seed, benchmark=benchmark, exp_name=exp_name, *kwargs)
File "/home/wenwen/workspace/HairFastGAN/utils/seed.py", line 28, in wraps
result = func(args, kwargs)
File "/home/wenwen/workspace/HairFastGAN/utils/time.py", line 34, in wraps
return func(args, kwargs)
File "/home/wenwen/workspace/HairFastGAN/hair_swap.py", line 51, in __swap_from_tensors
align_shape = self.align.align_images('face', 'shape', name_to_embed, kwargs)
File "/home/wenwen/workspace/HairFastGAN/scripts/blending_gen.py", line 35, in wrapper
return func(args, kwargs)
File "/home/wenwen/anaconda3/envs/hair/lib/python3.10/site-packages/torch/autograd/grad_mode.py", line 27, in decorate_context
return func(*args, *kwargs)
File "/home/wenwen/workspace/HairFastGAN/models/Alignment.py", line 130, in align_images
gen1_sean = decode_sean(self.sean_model, img1_code.unsqueeze(0), target_mask)
File "/home/wenwen/workspace/HairFastGAN/models/sean_codes/models/pix2pix_model.py", line 324, in decode_sean
generated = sean_model(data, mode='UI_mode')[0] # [3, 256, 256]
File "/home/wenwen/anaconda3/envs/hair/lib/python3.10/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl
return forward_call(input, kwargs)
File "/home/wenwen/workspace/HairFastGAN/models/sean_codes/models/pix2pix_model.py", line 45, in forward
input_semantics, real_image = self.preprocess_input(data)
File "/home/wenwen/workspace/HairFastGAN/models/sean_codes/models/pix2pix_model.py", line 133, in preprocess_input
if data['obj_dic'][str(idx)]['ACE'].device.type == 'cpu':
KeyError: 'ACE'
Process finished with exit code 1
As shown in the figure, further investigation revealed that after executing _img1_code, img2_code = encode_sean(self.sean_model, images, labels)_, the result in img1_code contains rows that are entirely zeros. This leads to a failure in passing the check in decodesean() during if not torch.all(cur_code == 0): obj_dic[str(idx)]['ACE'] = cur_code_.
Finally, an error occurs when running '_if data['obj_dic'][str(idx)]['ACE'].device.type == 'cpu':_.
How can I fix this bug?
Thanks!