Thank you for releasing the code.
I'm excited to work with it, but I've encountered some issues while trying to replicate the work. I noticed that in the datasets.py, there are still traces of personal file paths.
Additionally, the process for Data Preparation is not entirely clear. For instance, the method used for frame extraction from Kinetics frames is not specified, and it seems to differ from the approach used in VideoMAE. Could you provide a more detailed version of the data preparation steps? A clear and concise explanation would be greatly appreciated.
1162431418
commented
2 weeks ago
Hello,
Thank you for releasing the code. I'm excited to work with it, but I've encountered some issues while trying to replicate the work. I noticed that in the datasets.py, there are still traces of personal file paths.
Additionally, the process for Data Preparation is not entirely clear. For instance, the method used for frame extraction from Kinetics frames is not specified, and it seems to differ from the approach used in VideoMAE. Could you provide a more detailed version of the data preparation steps? A clear and concise explanation would be greatly appreciated.