To make the conda env, use:
conda env create -f environment.yml
After creating the conda, activate it using:
conda activate mlda
For recaptioning, get api key from https://aistudio.google.com/app/apikey and put it inside of .env as the following:
GEMINI_API_KEY=<YOUR_API_KEY>
Run the python file to run filtering and recaptioning:
python inference.py
Use config.yaml to change the batch to run.
Don't forget to run it from root folder ("./filtrain")
Use python utils/get_first_n_videos.py to get the first few videos for easier testing.
Video -> Clips -> Scene Split No Video: 18750 No Clips: 560755 No Scenes: 1504248
https://drive.google.com/drive/folders/1R52ueV6czG4M409YC6ca7W5QUqhCyoXg