Fast-RCNN target detection method based on FPGA

Abstract

The invention discloses an FPGA (Field Programmable Gate Array)-based master-RCNN (Recurrent Convolutional Neural Network) target detection method, which relates to the field of intelligent identification. The method comprises the following steps: preprocessing a data set; constructing a faster-RCNN model; loading the data set into the faster-RCNN model, and customizing an FPGA i according to thefaster-RCNN model; training a faster-RCNN model by using the customized FPGA; performing testing according to a training result of the faster-RCNN model, if the training result is lower than an average precision rate AP threshold, modifying parameters, performing training again, and testing the training result until the training result reaches a threshold; and inputting a picture to be detected, and performing target recognition by using the trained faster-RCNN model. According to the invention, each processing module of the FPGA is customized according to the faster-RCNN model; according to the invention, the method can achieve the precise recognition of an object through the Faster-RCNN model, also can solve a problem that the recognition speed of the Faster-RCNN model is slow, and achieves a higher detection speed, higher detection precision, better performances and lower power consumption.

Description

technical field [0001] The invention relates to the field of intelligent identification, in particular to an FPGA-based faster-RCNN target detection method. Background technique [0002] With the development of intelligent recognition technology, the terminal system needs to detect surrounding objects, especially in the field of automatic driving. For the sake of personal safety, target recognition must be fast and accurate. Therefore, it is of great practical significance to construct a fast and accurate object detection method. [0003] Existing target detection for autonomous driving requires both high precision and high speed. Existing target detection algorithms based on deep learning, such as SSD and YOLO, are fast but not accurate enough. In addition, the faster-RCNN algorithm is accurate enough but not fast enough in target detection. Each type of detection has more or less specific problems. For example, in the actual driving process, once the target detection is...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a faster-RCNN target detection method based on FPGA, based on the existing deep learning network and computer vision technology, and use FPGA to carry out deep customization according to the faster-RCNN model, FPGA can realize accurate Recognize objects, and can also solve the problem of slow recognition speed, realize parallel computing of the faster-RCNN model, and achieve the purpose of accelerating faster-RCNN target detection

Images