Teach-DETR: Better Training DETR with Teachers
This work addresses the challenge of enhancing accuracy and robustness in object detection for computer vision applications, representing an incremental improvement over existing DETR methods.
The paper tackles the problem of improving DETR-based object detectors by introducing Teach-DETR, a training scheme that uses predicted boxes from teacher detectors to transfer knowledge, resulting in a consistent performance boost, such as increasing mean average precision from 57.8% to 58.9% on the MSCOCO dataset.
In this paper, we present a novel training scheme, namely Teach-DETR, to learn better DETR-based detectors from versatile 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 with Swin-Large backbone, 4 scales of feature maps and 36-epoch training schedule, from 57.8% to 58.9% in terms of mean average precision on MSCOCO 2017 validation set. Code will be available at https://github.com/LeonHLJ/Teach-DETR.