search

ttengwang / PDVC

End-to-End Dense Video Captioning with Parallel Decoding (ICCV 2021)
MIT License
200 stars 23 forks source link

How to extract C3D features？ #54

Open Wangdanchunbufuz opened 7 months ago

Wangdanchunbufuz commented 7 months ago

Thank you for your excellent work. I want to run your model on my own data set, but you are training based on extracted features, can you disclose how to extract C3D video features from the data set? Thank you very much!

ttengwang commented 7 months ago

The C3D features are provided by the creators of ActivityNet Captions. For additional details, please visit http://activity-net.org/challenges/2016/download.html#c3d.

Wangdanchunbufuz commented 7 months ago

The C3D features are provided by the creators of ActivityNet Captions. For additional details, please visit http://activity-net.org/challenges/2016/download.html#c3d.

Thank you very much for your reply. I want to train this model on my own data set. In the first step, I follow your instructions to extract tsp features of the video. pkl and.npy files are obtained at the same time, but python merge_pkl_files_into_one_h5_feature_file.py is converted to.h5 using the Python merge_pkl_files_into_one_H5_feature_file. py according to the readme, but errors are reported during the conversion process. I cannot get the.h5 file. How should I proceed? image