RuntimeError: invalid argument 0: Sizes of tensors must match except in dimension 0. Got 940 and 1254 in dimension 2

[Jessicall]

Hi,good job! @AliaksandrSiarohin
My command is " CUDA_VISIBLE_DEVICES=0,1 python run.py --config config/fashion-256.yaml --device_ids 0,1"
My fashion-dataset folder is

（I don't process the data such as crop operation,）

And i get the following error:
Traceback (most recent call last):
File "run.py", line 81, in
train(config, generator, discriminator, kp_detector, opt.checkpoint, log_dir, dataset, opt.device_ids)
File "/remote-home/my/pycharmprojects/first-order-model/train.py", line 50, in train
for x in dataloader:
File "/usr/local/miniconda3/envs/animation1/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 637, in next
return self._process_next_batch(batch)
File "/usr/local/miniconda3/envs/animation1/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 658, in _process_next_batch
raise batch.exc_type(batch.exc_msg)
RuntimeError: Traceback (most recent call last):
File "/usr/local/miniconda3/envs/animation1/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 138, in _worker_loop
samples = collate_fn([dataset[i] for i in batch_indices])
File "/usr/local/miniconda3/envs/animation1/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 229, in default_collate
return {key: default_collate([d[key] for d in batch]) for key in batch[0]}
File "/usr/local/miniconda3/envs/animation1/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 229, in
return {key: default_collate([d[key] for d in batch]) for key in batch[0]}
File "/usr/local/miniconda3/envs/animation1/lib/python3.5/site-packages/torch/utils/data/dataloader.py", line 218, in default_collate
return torch.stack([torch.from_numpy(b) for b in batch], 0)
RuntimeError: invalid argument 0: Sizes of tensors must match except in dimension 0. Got 940 and 1254 in dimension 2 at /pytorch/aten/src/TH/generic/THTensorMoreMath.cpp:1333

[AliaksandrSiarohin]

Resize your videos to the same 256x256 size prior to training, as explained in README.md.

[Jessicall]

Thanks @AliaksandrSiarohin
And should i use the command "python crop-video.py --inp some_youtube_video.mp4"?
How can i crop all videos at the same time?

[AliaksandrSiarohin]

No. This script is not for resizing. Use xargs + ffmpeg for resizing all the videos.

[Jessicall]

@AliaksandrSiarohin
How to use the xargs when i run crop-video.py to process all videos?
Could you show me the command? I'm so confused,thannnnks

[AliaksandrSiarohin]

Crop-video.py script is only for faces. Fashion is full body dataset. Why you are using it there is not clear. You shold resize all the videos. Write a python script to do this or use xargs + ffmpeg.

[Jessicall]

@AliaksandrSiarohin
Oh i got it
And can you show me the command of 'xargs + ffmpeg'?
I'M a newbie hhh,thanks for your patience.

[AliaksandrSiarohin]

github issues is for reporting project specific bugs...

Use this python script to resize videos, adjust d and d_out appropriatly:

import imageio
import os
from tqdm import tqdm
from skimage.transform import resize

d = 'Videos_dataset_validation'
d_out = 'fashion-png/test'
os.makedirs(d_out)
for image in tqdm(os.listdir(d)):
    try:
       reader = imageio.get_reader(os.path.join(d, image))
       frames = []
       for frame in reader:
           frames.append(resize(frame, (256, 256)))
       reader.close()
    except:
       reader.close()
       None 
    os.makedirs(os.path.join(d_out, image))
    [imageio.imsave(os.path.join(d_out, image, str(i).zfill(7) + '.png'), frame) for i, frame in enumerate(frames)]

[Jessicall]

thanks for your help