Testing without Masks

[ghost]

Hello! And thank you for this latest work. I do apologize for this question if there is an easy answer that I have missed in the code. The inclusion of the evaluator is super helpful, but was curious if it was possible to amend the code to allow for testing when I do not have a mask of the image I wish to test, only the image, and still output the predicted mask?

[nnizhang]

You can run 'python train_test_eval.py --Training False --Testing True --Evaluation False' to predict the masks only from test images.

[ghost]

Thank you for the quick reply! and I do apologize for a follow up, when attempting to run that code with only the image in the test folder and no matching mask in the mask folder this is the error it throws:

adopt transformer encoder for tokens-to-token
Model loaded from ./pretrained_model/80.7_T2T_ViT_t_14.pth.tar
Model loaded from checkpoint/RGB_VST.pth

                   Starting testing:
                       dataset: DUTS
                       Testing size: 1
                   
/usr/local/lib/python3.7/dist-packages/torch/nn/functional.py:1806: UserWarning: nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.
  warnings.warn("nn.functional.sigmoid is deprecated. Use torch.sigmoid instead.")
dataset:DUTS, cost:70.75262069702148
eval[MAE]:DUTS/DUTS-TE dataset with RGB_VST method.
Traceback (most recent call last):
  File "train_test_eval.py", line 47, in <module>
    main.evaluate(args)
  File "/content/VST/RGB_VST/Evaluation/main.py", line 33, in evaluate
    print(thread.run())
  File "/content/VST/RGB_VST/Evaluation/evaluator.py", line 21, in run
    mae = self.Eval_mae()
  File "/content/VST/RGB_VST/Evaluation/evaluator.py", line 76, in Eval_mae
    for pred, gt in self.loader:
  File "/content/VST/RGB_VST/Evaluation/dataloader.py", line 31, in __getitem__
    gt = Image.open(self.label_path[item]).convert('L')
  File "/usr/local/lib/python3.7/dist-packages/PIL/Image.py", line 2843, in open
    fp = builtins.open(filename, "rb")
FileNotFoundError: [Errno 2] No such file or directory: './Data/DUTS/DUTS-TE/DUTS-TE-Mask/01.png'

I am hoping to find a way to remove that dependency, as I don't have the mask for the image I wish to test. Possible?

[ghost]

Actually, follow on, adding the training flag to that string causes this error, removing it causes the error in the first comment. Including the flag gets this error, which makes sense, I don't have any images in the training folders.

adopt transformer encoder for tokens-to-token
Model loaded from ./pretrained_model/80.7_T2T_ViT_t_14.pth.tar
        Starting training:
            Train steps: 60000
            Batch size: 11
            Learning rate: 0.0001
            Training size: 0
        
Starting epoch 1/200.
epoch:1-------lr:0.0001
Traceback (most recent call last):
  File "train_test_eval.py", line 43, in <module>
    Training.train_net(num_gpus=num_gpus, args=args)
  File "/content/VST/RGB_VST/Training.py", line 46, in train_net
    mp.spawn(main, nprocs=num_gpus, args=(num_gpus, args))
  File "/usr/local/lib/python3.7/dist-packages/torch/multiprocessing/spawn.py", line 230, in spawn
    return start_processes(fn, args, nprocs, join, daemon, start_method='spawn')
  File "/usr/local/lib/python3.7/dist-packages/torch/multiprocessing/spawn.py", line 188, in start_processes
    while not context.join():
  File "/usr/local/lib/python3.7/dist-packages/torch/multiprocessing/spawn.py", line 150, in join
    raise ProcessRaisedException(msg, error_index, failed_process.pid)
torch.multiprocessing.spawn.ProcessRaisedException: 

-- Process 0 terminated with the following error:
Traceback (most recent call last):
  File "/usr/local/lib/python3.7/dist-packages/torch/multiprocessing/spawn.py", line 59, in _wrap
    fn(i, *args)
  File "/content/VST/RGB_VST/Training.py", line 176, in main
    print('Epoch finished ! Loss: {}'.format(epoch_total_loss / iter_num))
ZeroDivisionError: division by zero

[ghost]

Annnnnnnd you can just ignore, and i'll close this. It appears only having !python train_test_eval.py --Testing True, works. including any of the other flags causes weirdness.

All good!