This section hosts the detail information to conduct narration file pre-processing for model pre-training.
First get official Egoclip file and the annotated training split files from visual/moment/natural language queries challenges.
python create_pretrain_data_from_egoclip.py
python filter_missing_video.py
The first code aims to generate the pretrained data, and the second code aims to filter some train samples.