A Lightweight YOLOv5 Optimization of Coordinate Attention
Open Access
- 30 January 2023
- journal article
- research article
- Published by MDPI AG in Applied Sciences
- Vol. 13 (3) , 1746
Abstract
As Machine Learning technologies evolve, there is a desire to add vision capabilities to all devices within the IoT in order to enable a wider range of artificial intelligence. However, for most mobile devices, their computing power and storage space are affected by factors such as cost and the tight supply of relevant chips, making it impossible to effectively deploy complex network models to small processors with limited resources and to perform efficient real-time detection. In this paper, YOLOv5 is studied to achieve the goal of lightweight devices by reducing the number of original network channels. Then detection accuracy is guaranteed by adding a detection head and CA attention mechanism. The YOLOv5-RC model proposed in this paper is 30% smaller and lighter than YOLOv5s, but still maintains good detection accuracy. YOLOv5-RC network models can achieve a good balance between detection accuracy and detection speed, with potential for its widespread use in industry.Keywords
Funding Information
- National Natural Science Foundation of China (61602161, 61772180, 2020BAB012, 2021046)
- Hubei Province Science and Technology (61602161, 61772180, 2020BAB012, 2021046)
- Hubei University of Technology (61602161, 61772180, 2020BAB012, 2021046)
This publication has 29 references indexed in Scilit:
- ShuffleNet V2: Practical Guidelines for Efficient CNN Architecture Design2018
- Mask R-CNNIEEE Transactions on Pattern Analysis and Machine Intelligence, 2018
- ShuffleNet: An Extremely Efficient Convolutional Neural Network for Mobile Devices2018
- YOLO9000: Better, Faster, Stronger2017
- SSD: Single Shot MultiBox Detector2016
- Faster R-CNN: Towards Real-Time Object Detection with Region Proposal NetworksIEEE Transactions on Pattern Analysis and Machine Intelligence, 2016
- You Only Look Once: Unified, Real-Time Object Detection2016
- Fast R-CNN2015
- Region-Based Convolutional Networks for Accurate Object Detection and SegmentationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2015
- Spatial Pyramid Pooling in Deep Convolutional Networks for Visual RecognitionIEEE Transactions on Pattern Analysis and Machine Intelligence, 2015