Teach-DETR: Better Training DETR With Teachers
Publication in refereed journal
全文
替代計量分析
.
其它資訊
摘要In this paper, we present a novel training scheme, namely Teach-DETR, to better train DETR-based detectors from versatile types of teacher detectors. We show that the predicted boxes from teacher detectors are effective medium to transfer knowledge of teacher detectors, which could be either RCNN-based or DETR-based detectors, to train a more accurate and robust DETR model. This new training scheme can easily incorporate the predicted boxes from multiple teacher detectors, each of which provides parallel supervisions to the student DETR. Our strategy introduces no additional parameters and adds negligible computational cost to the original detector during training. During inference, Teach-DETR brings zero additional overhead and maintains the merit of requiring no non-maximum suppression. Extensive experiments show that our method leads to consistent improvement for various DETR-based detectors. Specifically, we improve the state-of-the-art detector DINO Zhang et al. 2022 with Swin-Large Liu et al. 2021 backbone, 4-scale feature pyramid and 36-epoch training schedule, from 57.8% to 58.9% in terms of mean average precision on COCO 2017 val set.
著者Linjiang Huang, Kaixin Lu, Guanglu Song, Liang Wang, Si Liu, Yu Liu, Hongsheng Li
期刊名稱IEEE Transactions on Pattern Analysis and Machine Intelligence
出版年份2023
月份12
卷號45
期次12
出版社IEEE Computer Society
頁次15759 - 15771
國際標準期刊號0162-8828
語言美式英語
