ValueError: Target size must be the same as input size

[Zy1013921842]

🐛 Bug

When I run the train_net.py, it raises a ValueError: Target size (torch.Size([2, 28, 28])) must be the same as input size (torch.Size([4, 28, 28])). The traceback is below:

Traceback (most recent call last):
File "tools/train_net.py", line 186, in
main()
File "tools/train_net.py", line 179, in main
model = train(cfg, args.local_rank, args.distributed)
File "tools/train_net.py", line 85, in train
arguments,
File "C:\Users\opt\maskrcnn-benchmark\maskrcnn_benchmark\engine\trainer.py", line 67, in do_train
loss_dict = model(images, targets)
File "D:\zhaoyang\anaconda\anaconda\envs\maskrcnn_benchmark\lib\site-packages\torch\nn\modules\module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "C:\Users\opt\apex\apex\amp_initialize.py", line 194, in new_fwd
**applier(kwargs, input_caster))
File "C:\Users\opt\maskrcnn-benchmark\maskrcnn_benchmark\modeling\detector\generalized_rcnn.py", line 52, in forward
x, result, detector_losses = self.roi_heads(features, proposals, targets)
File "D:\zhaoyang\anaconda\anaconda\envs\maskrcnn_benchmark\lib\site-packages\torch\nn\modules\module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "C:\Users\opt\maskrcnn-benchmark\maskrcnn_benchmark\modeling\roi_heads\roi_heads.py", line 39, in forward
x, detections, loss_mask = self.mask(mask_features, detections, targets)
File "D:\zhaoyang\anaconda\anaconda\envs\maskrcnn_benchmark\lib\site-packages\torch\nn\modules\module.py", line 493, in call
result = self.forward(*input, **kwargs)
File "C:\Users\opt\maskrcnn-benchmark\maskrcnn_benchmark\modeling\roi_heads\mask_head\mask_head.py", line 77, in forward
loss_mask = self.loss_evaluator(proposals, mask_logits, targets)
File "C:\Users\opt\maskrcnn-benchmark\maskrcnn_benchmark\modeling\roi_heads\mask_head\loss.py", line 126, in call
mask_logits[positive_inds, labels_pos], mask_targets
File "D:\zhaoyang\anaconda\anaconda\envs\maskrcnn_benchmark\lib\site-packages\torch\nn\functional.py", line 2161, in binary_cross_entropy_with_logits
raise ValueError("Target size ({}) must be the same as input size ({})".format(target.size(), input.size()))
ValueError: Target size (torch.Size([2, 28, 28])) must be the same as input size (torch.Size([4, 28, 28]))

By the way, before I got this error, actually I got another IndexError: list index out of range in segmentation_mask.py, whose traceback is below. I noticed that #725 mentioned a similar bug as mine, but in their case, they were using pytorch=1.2 while the version I was using is pytorch=1.1. Anyway, I tried their method which is replacing all torch.uint8 with torch.bool in segmentation_mask.py but it didn't work. Then I found another issue #704 which says it might be the problem of the mismatch of size between item and self.polygons. At last I still didn't konw how to solve this so I just bypassed the problem by using try-except, I mention this here because I don't know whether this trick has anything to do with the above valueError.

Traceback (most recent call last):
File "tools/train_net.py", line 186, in
main()
File "tools/train_net.py", line 179, in main
model = train(cfg, args.local_rank, args.distributed)
File "tools/train_net.py", line 85, in train
arguments,
File "C:\Users\opt\maskrcnn-benchmark\maskrcnn_benchmark\engine\trainer.py", line 57, in do_train
for iteration, (images, targets, _) in enumerate(data_loader, start_iter):
File "D:\zhaoyang\anaconda\anaconda\envs\maskrcnn_benchmark\lib\site-packages\torch\utils\data\dataloader.py", line 568, in next
return self._process_next_batch(batch)
File "D:\zhaoyang\anaconda\anaconda\envs\maskrcnn_benchmark\lib\site-packages\torch\utils\data\dataloader.py", line 608, in _process_next_batch
raise batch.exc_type(batch.exc_msg)
IndexError: Traceback (most recent call last):
File "D:\zhaoyang\anaconda\anaconda\envs\maskrcnn_benchmark\lib\site-packages\torch\utils\data_utils\worker.py", line 99, in _worker_loop
samples = collate_fn([dataset[i] for i in batch_indices])
File "D:\zhaoyang\anaconda\anaconda\envs\maskrcnn_benchmark\lib\site-packages\torch\utils\data_utils\worker.py", line 99, in
samples = collate_fn([dataset[i] for i in batch_indices])
File "C:\Users\opt\maskrcnn-benchmark\maskrcnn_benchmark\data\datasets\coco.py", line 91, in getitem
target = target.clip_to_image(remove_empty=True)
File "C:\Users\opt\maskrcnn-benchmark\maskrcnn_benchmark\structures\bounding_box.py", line 223, in clip_to_image
return self[keep]
File "C:\Users\opt\maskrcnn-benchmark\maskrcnn_benchmark\structures\bounding_box.py", line 208, in getitem
bbox.add_field(k, v[item])
File "C:\Users\opt\maskrcnn-benchmark\maskrcnn_benchmark\structures\segmentation_mask.py", line 528, in getitem
selected_instances = self.instances.getitem(item)
File "C:\Users\opt\maskrcnn-benchmark\maskrcnn_benchmark\structures\segmentation_mask.py", line 433, in getitem
selected_polygons.append(self.polygons[i])
IndexError: list index out of range

Here is my environment:

• PyTorch Version: 1.1:
• OS: Windows10
• Python version: python3.6
• CUDA/cuDNN version: CUDA=10.0, cuDNN=10.1

At the end, I wanna mention another thing. I managed to train another sets of images which are gray pictures but this time the images I want to train are all RGB pictures, I don't know whether this makes any difference.

Any help is appreciated!

[nancymao]

Hi, have you solved the problem? I'm using torch-nightling 1.0.0 and having same problem as you. PLEASE let me known if u solve the problem! THX !!

[Zy1013921842]

@nancymao
Hello, sorry to reply late. In my case accually I got this problem after doing some unsuitable changes in the code to solve another error: IndexError: list index out of range. You can see#704 as a refference, I'm not sure that you are in the same situation as I was. Good luck!