Enhancing Source-Free Domain Adaptive Object Detection with Low-confidence Pseudo Label Distillation

TL;DR

This work tackles SFOD by addressing the bias of relying solely on High-confidence Pseudo Labels (HPL) in Mean-Teacher self-training. It introduces Low-confidence Pseudo Label Distillation (LPLD), which mines hard positives from Region Proposal Network proposals not overlapped with HPL and refines them through a KL-divergence loss weighted by teacher-student feature distance. The approach yields state-of-the-art or competitive results across multiple cross-domain benchmarks (Cityscapes→Foggy Cityscapes, Sim10k→Cityscapes, KITTI→Cityscapes, Pascal-VOC→Clipart/Watercolor), notably reducing false negatives for small and hard-to-detect objects. By leveraging domain-invariant cues from the source model and explicitly exploiting low-confidence proposals, LPLD enhances adaptation while maintaining stability and avoiding source-data reliance, with code released for reproducibility.

Abstract

Source-Free domain adaptive Object Detection (SFOD) is a promising strategy for deploying trained detectors to new, unlabeled domains without accessing source data, addressing significant concerns around data privacy and efficiency. Most SFOD methods leverage a Mean-Teacher (MT) self-training paradigm relying heavily on High-confidence Pseudo Labels (HPL). However, these HPL often overlook small instances that undergo significant appearance changes with domain shifts. Additionally, HPL ignore instances with low confidence due to the scarcity of training samples, resulting in biased adaptation toward familiar instances from the source domain. To address this limitation, we introduce the Low-confidence Pseudo Label Distillation (LPLD) loss within the Mean-Teacher based SFOD framework. This novel approach is designed to leverage the proposals from Region Proposal Network (RPN), which potentially encompasses hard-to-detect objects in unfamiliar domains. Initially, we extract HPL using a standard pseudo-labeling technique and mine a set of Low-confidence Pseudo Labels (LPL) from proposals generated by RPN, leaving those that do not overlap significantly with HPL. These LPL are further refined by leveraging class-relation information and reducing the effect of inherent noise for the LPLD loss calculation. Furthermore, we use feature distance to adaptively weight the LPLD loss to focus on LPL containing a larger foreground area. Our method outperforms previous SFOD methods on four cross-domain object detection benchmarks. Extensive experiments demonstrate that our LPLD loss leads to effective adaptation by reducing false negatives and facilitating the use of domain-invariant knowledge from the source model. Code is available at https://github.com/junia3/LPLD.

Figures (43)

• Figure 1: (a) 2D Histogram of Proposals on Cityscapescityscapes to Foggy Cityscapesfoggy. Confidence Score and IoU with Ground Truth illustrates that before adaptation, the source-trained model often overlooks hard positive objects in the proposals with high IoUs but low confidence scores (red boxes). After adaptation, our LPLD loss promotes the detector to effectively capture hard positives with high-confidence scores in comparison to the Mean-Teacher (MT) mean_teacher based SFOD model which utilizes only the High-confidence Pseudo Labels (HPL). (b) False Negative Rate (FNR) per Training Epoch. Our model shows a consistently lower FNR than the MT baseline on hard-positive objects (e.g., Minor classes, Small objects).
• Figure 1: 2D Histogram of Proposals on Cityscapes $\rightarrow$ Foggy Cityscapes.
• Figure 2: Comparison of two pseudo label supervision paradigms in Mean-Teacher SFOD methods. (a) High-confidence Pseudo Labels (HPL), which are provided as supervision for localization and classification to the student detector. (b) Low-confidence Pseudo Labels (LPL) supervise the student through our sophisticated distillation process.
• Figure 2: Number of instances and AP50 gain for each category.
• Figure 3: The overview of the proposed adaptive LPL Distillation framework.