hello,i train your data,but when i want to test the model,AttributeError: 'NoneType' object has no attribute 'seek'. You can only torch.load from a file that is seekable. Please pre-load the data into a buffer like io.BytesIO and try to load from it instead. about cascadepsp

[hkchengrex]

What command did you use?

[zhujiesuper]

python eval.py --dataset CascadePSP-master/testset_directory --CascadePSP-master/weights/3_2020-05-17_12:15:38/model_45705 --output CascadePSP-master/testset_directory --------the model trained by your method completely

[hkchengrex]

The "--model" phrase is missing before the model file.

[zhujiesuper]

Thank YOU! I change this,but A new broken,0 images found,,,my code is python eval.py --dataset /home/zj/PycharmProjects/CascadePSP-master/CascadePSP-master/testset_directory --model /home/zj/PycharmProjects/CascadePSP-master/CascadePSP-master/weights/3_2020-05-17_12:15:38/model_45000 --output /home/zj/PycharmProjects/CascadePSP-master/CascadePSP-master/output_directory
and the testset_directory I put xxxx_gt.png,xxxx_im.png,xxx_seg.png which come from the BIG dataset

[hkchengrex]

The image files should be in .jpg. If your images are in .png you can modify the dataset class to take png pretty easily.

[zhujiesuper]

ThankYou！I solve this by change --dataset to --dir ,and png is ok ,jpg need modify! see the result,it is an awesome job!

[zhujiesuper]

but why in eval.py i need to put the _gt.png in test,did the _gt.png affect the result?

[hkchengrex]

Thanks for pointing that out! We actually had a typo in the document which we will now fix. The _gt files do not affect the results. They are just copied to the output folder for future evaluation.

[zhujiesuper]

Nothing,Nothing,This job is very awesome!

[hkchengrex]

Thanks! 😄