SCL: Towards Accurate Domain Adaptive Object Detection via Gradient Detach Based Stacked Complementary Losses

TL;DR

This paper introduces SCL, a novel gradient detach based stacked complementary losses method that enhances unsupervised domain adaptive object detection by learning more discriminative features through auxiliary losses and gradient detach training, outperforming existing methods.

Contribution

Proposes a new SCL method that leverages auxiliary losses with gradient detach to improve domain adaptive object detection, addressing loss interaction issues in prior methods.

Findings

Achieves 37.9% mAP from Cityscapes to FoggyCityscapes, surpassing previous methods.

Demonstrates significant performance improvements across seven datasets.

Validates the effectiveness of gradient detach and auxiliary losses in domain adaptation.

Abstract

Unsupervised domain adaptive object detection aims to learn a robust detector in the domain shift circumstance, where the training (source) domain is label-rich with bounding box annotations, while the testing (target) domain is label-agnostic and the feature distributions between training and testing domains are dissimilar or even totally different. In this paper, we propose a gradient detach based stacked complementary losses (SCL) method that uses detection losses as the primary objective, and cuts in several auxiliary losses in different network stages accompanying with gradient detach training to learn more discriminative representations. We argue that the prior methods mainly leverage more loss functions for training but ignore the interaction of different losses and also the compatible training strategy (gradient detach updating in our work). Thus, our proposed method is a more syncretic adaptation learning process. We conduct comprehensive experiments on seven datasets, the results demonstrate that our method performs favorably better than the state-of-the-art methods by a significant margin. For instance, from Cityscapes to FoggyCityscapes, we achieve 37.9% mAP, outperforming the previous art Strong-Weak by 3.6%.