search

cosmaadrian / multimodal-depression-from-video

Official source code for the paper: "Reading Between the Frames Multi-Modal Non-Verbal Depression Detection in Videos"
Other
47 stars 7 forks source link

correct structure to store the videos #87

Open palashmoon opened 6 months ago

palashmoon commented 6 months ago

HI @david-gimeno.. I have downloaded all the videos from the D-vlog dataset.. should i split it based on the ids given in the test, train and validation.csv file?? or there is a separate file python3 ./scripts/feature_extraction/dvlog/extract_wavs.py --csv-path ./data/D-vlog/video_ids.csv --column-video-id video_id --video-dir $VIDEO_DIR --dest-dir $WAV_DIR video_id.csv which is missing in the repo?? please help me with this..

david-gimeno commented 4 months ago

According to the script you refer to, all the video clips composing D-Vlog should be stored in the same directory. Feel free to modify the script to use it based on your own preferences. You can use the package glob to avoid compounding the pathways, etc.

As you can see here in the code implementing our dataset, which it is the object in charge of providing the data to train and evaluate the final model, we will later have CSVs defining our dataset splits (train, val, test), but the videos and features can be stored all together in the same directory.