This project demonstrates polyinfer's ability to orchestrate inference across 3 different compute devices (GPU, CPU, and NPU) within a single real-time video pipeline.