Multi-Camera Person Re-Identification System

An Ontario hospital is seeking a vendor to provide a secure, real-time multi-camera person re-identification system for staff tracking within clinical environments. The system should leverage advanced deep learning models and edge-based deployment to support workflow analytics while maintaining strict privacy compliance standards.

Key requirements include the development of a real-time ingestion and inference pipeline for person detection and tracking across multiple cameras, such as fisheye and bullet cameras. The solution should feature a fine-tuned deep learning pipeline (e.g., OSNet or similar) trained on hospital-specific data to optimize performance under clinical conditions. Consistent identity management must be achieved through a robust multi-camera matching mechanism.

Privacy is paramount; therefore, no original frames may be stored, and all embeddings must be encrypted. Deployment is to be performed on hospital-owned edge devices such as NVIDIA AGX Orin, ensuring low-latency performance. The selected vendor will also be responsible for providing comprehensive documentation and APIs to facilitate integration with workflow analytics tools. All vendor inquiries must be submitted by December 22, 2025.