CoolFace
Apppublic

alinawazmahar/cctv-visual-rag

sourceHugging Facemitupdated 3mo agoView on Hugging Face
2likes
App README

๐ŸŽฅ CCTV Visual RAG

Search through surveillance / CCTV video frames using natural language queries. No manual scrubbing. Just type what you're looking for.

How it works

Video โ†’ Frame Sampling (1 FPS) โ†’ CLIP Embeddings โ†’ FAISS Index
Query โ†’ CLIP Text Embedding โ†’ Similarity Search โ†’ Top-K Frames + Timestamps

Stack

ComponentTool
Embeddingsopenai/clip-vit-base-patch32
Vector StoreFAISS (IndexFlatIP)
Frame ExtractionOpenCV
UIGradio

Example Queries

  • โ€”red car
  • โ€”person walking
  • โ€”empty road
  • โ€”multiple vehicles

Author

Built by Ali Nawaz โ€” CS Student

HuggingFace ยท GitHub ยท LinkedIn