Leveraging Confident Image Regions for Source-Free Domain-Adaptive Object Detection

2025-01-17

AI Summary

This paper addresses source-free domain adaptive object detection (SF-DAOD), where a pre-trained detector must be adapted to an unlabeled target domain without access to source-domain data. To tackle this challenge, we propose the first region-level data augmentation paradigm tailored for source-free settings: leveraging a teacher-student collaborative framework, it automatically crops high-confidence object regions and reconstructs images with semantically consistent augmentations using reliable pseudo-labels. Furthermore, we design a source-free domain adaptation training strategy to mitigate pseudo-label noise and domain shift. Our method is evaluated on three cross-domain traffic scene benchmarks, achieving new state-of-the-art performance on two of them and significantly outperforming existing source-free adaptive approaches.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Application Domains: Transportation

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Source-free domain-adaptive object detection is an interesting but scarcely addressed topic. It aims at adapting a source-pretrained detector to a distinct target domain without resorting to source data during adaptation. So far, there is no data augmentation scheme tailored to source-free domain-adaptive object detection. To this end, this paper presents a novel data augmentation approach that cuts out target image regions where the detector is confident, augments them along with their respective pseudo-labels, and joins them into a challenging target image to adapt the detector. As the source data is out of reach during adaptation, we implement our approach within a teacher-student learning paradigm to ensure that the model does not collapse during the adaptation procedure. We evaluated our approach on three adaptation benchmarks of traffic scenes, scoring new state-of-the-art on two of them.

Problem

• Transfer Learning
• Object Detection
• Adaptation

Innovation

• Domain Adaptation
• Object Recognition
• Data Augmentation