FPGA Based Hardware Accelerated Object Tracking

Abstract—Object tracking is used in various day to day applications such as face recognition devices, CCTV camera applications, vehicle tracking systems. This paper presents the implementation of Tiny-Yolo V3 algorithm on the PYNQ-Z2 FPGA board for achieving hardware accelerated detection model for images and video inputs. Here we make use of the both software and hardware components and integrate them for optimal results. Initially the model was implemented and tested on CPU only model, then integrated hardware-assisted control modules on the FPGA fabric to support real-time operation. The proposed work achieves effective object detection with low power consumption. Future works will include DPU based full-YOLO network deployment. Keywords—FPGA, Tiny-Yolo V3, Object tracking, PYNQ-Z2, Hardware Acceleration, DPU