BROKI is an virality/engagement detection system that is designed to assist parents and caregivers in curating online content for children. It reviews media from a provided source, predicts the intensity and areas of the brain that light up, renders warnings for highly stimulating content with engagement/arousal frames report beside the content playback, and only escalates to a VLM for content understanding when the content appears unusually engaging. If the VLM also finds brainrot-style risk patterns, BROKI warns the caregiver and learns from a simple binary approve/disapprove action.

Built With

  • fastapi
  • ffprobe-python
  • httpx
  • mutagen
  • numpy
  • pillow
  • playwright
  • plotbrain-(fsaverage5)
  • pydantic
  • pydantic-settings
  • pytest
  • python-3.11+
  • ruff
  • scikit-learn
  • scipy
  • sqlite
  • tribe-v2-predictor
  • uvicorn
Share this project:

Updates

Submission history