Efficient Urban Object Detection on Edge Devices: A Study on Raspberry Pi ۴ and Intel NCS۲

Real-time object detection in urban environments plays an essential role, especially in applications such as traffic management and pedestrian safety. However, deploying deep neural object detectors on embedded devices is still challenging due to their computational constraints and power limitations. To address this, in this paper, we explore the integration of Raspberry Pi ۴ and Intel Movidius Neural Compute Stick ۲ (NCS۲) for real-time urban object detection. By utilizing the OpenVino-optimized person-vehicle-bike-detection-۲۰۰ model, we achieve an efficient balance between accuracy and processing speed. Our experimental results demonstrate the effectiveness of NCS۲ with Raspberry Pi ۴ in detecting urban objects. This integration increases processing speed while maintaining low energy consumption which makes it a good choice for urban monitoring on resource-limited devices. Furthermore, all of our implementation codes and setups are available on GitHub: https://github.com/hamed-tgh/Urban_object_detection_NCS۲_Raspberry

کلیدواژه ها:

Object detection ، Raspberry Pi ، Neural compute stick ، Deep learning