A Simultaneous Hole, Groove and Defect Detection Method Based on Yolov4-tiny and MMOE
Lists
Only you see this
Abstract
The quality detection of furniture boards is one of the most important parts of the furniture board quality control process. The boards may contain round holes, grooves, and defects for various reasons. In industry, it is necessary to detect them in real time. This paper proposes a multi-objective detection method based on the Yolov4-tiny and Multi-Gate Mixture-of-Experts(MMOE) structure that can simultaneously detect round holes, grooves and defects in high resolution board images in real time. Considering the high resolution of the board image and the small size of the dataset, data pre-processing and data augmentation are performed to reduce the size of the image and increase the size of the dataset. In consideration of the difference in the characteristics of the hole, groove and defect. the task of detecting defects and the task of detecting holes and grooves are decoupled, and different task headers and gate networks are designed. Finally, it is tested on a real production line and the combined mAP reaches 75.49% and the detection takes only 20ms.
BibTeX
@inproceedings{lin2022simultaneous,
author = {Haicheng Lin and Jie Gong and Xiang Chen and Xiaohua Xie and Hengliang Tang and Junfeng Jiang},
title = {{A Simultaneous Hole, Groove and Defect Detection Method Based on Yolov4-tiny and MMOE}},
booktitle = {IEEE/CVF International Conference on Computer Vision},
year = {2022},
doi = {10.1109/cvidliccea56201.2022.9825359},
}Loading related content...