Person/Object tracking in a video stream

Project description

The goal of the project was to create a person, face, and object and a tracking system for a low-resolution video stream. Our solution was integrated into the client’s security system to prevent shoplifting, monitor employees and improve general security.

Fine-tuned triplet based face representation (embedding) on the domain-specific dataset.

Technologies:

• PyTorch based face detection and vectorization.
• Faiss for fast vector search.
• MongoDB, Postgres for meta-data.