Target Detection and Recognition System Based on PYNQ Platform

Target detection and recognition is a very important part in artificial intelligence field. In recent years, how to achieve target detection with low power consumption has become a research hotspot. In this paper, based on the PYNQ platform, the neural network single step multi-frame detection algorithm (SSD) is adopted, the overall function planning and system construction are carried out through software and hardware cooperation. Firstly, the convolutional neural network (CNN) training model is used for training and then weight parameter files is generated. After that we use HLS to design custom IP core and build system access. The last step is calling PC program. The test results under Jupyter Notebook environment show that the system can achieve effective target classification with low power consumption, which proves the feasibility of the whole system design.