Lightweight pedestrian and vehicle detection method based on improved YOLO v4

Abstract

The invention discloses a lightweight pedestrian and vehicle detection method based on improved YOLO v4. The method comprises the following steps: 1) dividing a data set into a training set, a verification set and a test set; 2) performing clustering analysis on a real target frame in the data set label to obtain a priori frame size; 3) building a backbone network by using a Ghost module for extracting visual features of the data set, and reducing parameter quantity while ensuring feature quality; 4) improving the multi-scale sensing capability of the network by using hole convolution with different expansion rates; 5) carrying out feature aggregation on the extracted features; 6) inputting the aggregated features into a detection head, predicting the position and category of the target, and training a model by using a loss function; and 7) inputting a test set image into the network model obtained by training for detection, and outputting a target detection result. According to the method, the pedestrian and vehicle detection precision is guaranteed, meanwhile, the model parameter quantity is small, the detection speed is high, and the requirement for the performance of hardware equipment is lowered.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a lightweight pedestrian vehicle detection method based on improved YOLO v4. Background technique [0002] With the rapid development of artificial intelligence, as a branch of computer vision, object detection technology has achieved many breakthroughs. Thanks to technological breakthroughs, target detection technology has gradually moved towards practical applications, and has been widely used in many fields such as automatic driving, video surveillance, and national defense and military affairs. For autonomous driving, quickly and accurately identifying pedestrians and vehicles is an important part of ensuring the safety of autonomous driving. Although the current pedestrian and vehicle detection technology has made great progress, there are still some problems. First of all, for autonomous driving traffic scenarios, more target detection algorithms need to be deployed on edge...

AI Technical Summary

Problems solved by technology

With the continuous development of technology, many existing target detection algorithms already have high detection accuracy, but they are large in size and are not suitable for edge devices such as vehicle systems, and there is still room for improvement in detection speed

Images