Group-of-Picture Mode Acceleration for Efficient Object Detection in Video Streams

Although modern object detection AI models have the potential to be widely used in various applications such as autonomous vehicles, these are very computationally demanding. Using high-resolution image data further increases computational burden. Hence, we propose an acceleration method called Group of Picture (GoP) mode for video sequences by removing temporal redundancy, unlike existing model compression schemes. A GoP structure is composed only one key frame and several non-key frames. In GoP-mode, adopted frames only, while tracking employed predict position each following based on tracked trajectory momentum object. By using proposed method, thrilling latency saving can result multiple times execution speed so that both high accuracy obtained. theory, if adopt a with N frames, accelerated (1+N) equipping GoP-mode. The effect number variation detector equipped GoP-mode has been analyzed. According experimental results, mean average precision (mAP) adopting four competitive all Meanwhile, rate increased from original 8 per second (FPS) 35.8 FPS mobile platform-Jetson Nano, i.e. speedup 348%.