Error in loading videos for testing

[mr17m]

Hi, I trained my video classification model using the trainer of this repository and I used the annotation files (csv files) in data_list folders but because I only used limited number of video classes for training I removed other annotations for other classes in the csv files and I renamed the classes from 0 to N and also I was careful to apply such changes in the "kinetics_400_categroies.txt". Plus, I extracted the videos from their class folders and I put all the videos in a specific folder for testing and also training.

I used this argument for the train "_C.DATA.PATH_LABEL_SEPARATOR = "," " and no error appeared. So I use the same for the test but I see the following error:

Any idea what is wrong?

[Andy1621]

Can you try to print the label and see its type? It is possible the label is an empty string, because there is some error in you annotation file.

[mr17m]

@Andy1621 I solved the previous issue but I still face a new error. The question is updated with the new error. It is worthy to mention that I use the csv files in the repository but I only removed other classes in the csv files because I trained and want to test on only 25 classes. What is wrong?

[Andy1621]

There are some errors when loading the video. You can print the path of the video and check whether it is correct.

BTW, for an issue, you can post a new bug instead of changing the previous content, thus others can follow your issues.

[mr17m]

@Andy1621 , Thanks for the comment. I printed out the path of the videos but it is correct. Do you have any idea what is wrong?

[Andy1621]

Sorry for the late response! You can try to set num_worker=0, thus it may present the correct error message.

[mr17m]

I did but I faced with the following error:

Any idea?

[Andy1621]

Check the log, and you can find persistent_workers should be False when num_workers is 0. And set the corresponding hyperparameters in line158 in loader.py.