国产欧美精品一区二区,中文字幕专区在线亚洲,国产精品美女网站在线观看,艾秋果冻传媒2021精品,在线免费一区二区,久久久久久青草大香综合精品,日韩美aaa特级毛片,欧美成人精品午夜免费影视

Aiming at the requirements of the embedded real-time application of image object detection, the deep learning object detection algorithm based on MobileNet and Single Shot Multi-Box Detector (SSD) is optimized by the method of combining computational layers, and the real-time image object detection system is designed by using software and hardware combination method based on ZYNQ scalable processing platform. In the system, a multi-processor core deep learning algorithm accelerator is designed according to the optimized algorithm, and the software of the system is designed by PYTHON language. After several experiments, the processing speed of deep learning object detection system can reach 45 FPS, which is 4.9X faster than deep learning framework running on CPU and 1.7X faster than on GPU. It fully meets the requirements of real-time image object detection.