Uncertainty and eval problems

[wzh506]

Hi~, thank you for your great work,but I have trouble in trevi dataset, first , I find type(sampling_mask) is: <class 'NoneType'>,epochs now is 2000,is there something wrong? second,when I enter func eval_few_custom

  File "/data/wangzhaohui/github/wild-gaussians/wildgaussians/train.py", line 360, in train_command
    eval_few_custom(method, logger, train_dataset_eval_few, split="train", step=step, evaluation_protocol=evaluation_protocol,output_path=output_path)
    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/data/wangzhaohui/github/wild-gaussians/wildgaussians/train.py", line 165, in eval_few_custom
    k + "-nopt": v for k, v in compute_metrics(image_to_srgb(result_no_optim["color"], dtype=np.float32), images_f[i]).items()
                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/data/wangzhaohui/github/wild-gaussians/wildgaussians/evaluation.py", line 343, in compute_metrics
    mse_ = mse(pred, gt)
           ^^^^^^^^^^^^^
  File "/data/wangzhaohui/github/wild-gaussians/wildgaussians/evaluation.py", line 210, in mse
    assert a.shape == b.shape, f"Images must have the same shape, got {a.shape} and {b.shape}"
           ^^^^^^^^^^^^^^^^^^
AssertionError: Images must have the same shape, got (735, 507, 3) and (735, 508, 3)

it seems that result_no_optim["color"] face some problem. Thank you!

[jkulhanek]

Can you please pull the newest version? I've updated today. Or do you have the newest version?

[jkulhanek]

Sampling mask is expected to be none.