Domain Adaptive Object Detection via Dual-Stream Bilevel-Cycle Optimization

2026-06-30

AI Summary

This work addresses key challenges in unsupervised domain adaptive object detection—namely, unreliable pseudo-labels, training instability, and performance degradation caused by excessively large regression coordinate ranges. To overcome these limitations, the paper introduces a dual-stream, two-stage cyclic self-training (CST) framework that, for the first time, integrates CST into both classification and regression branches. By synergistically combining a Mean Teacher architecture with regression coordinate normalization, the proposed method effectively mitigates loss explosion and enhances cross-domain generalization. Theoretical analysis establishes rigorous regression bounds, and extensive experiments across four standard cross-domain benchmarks demonstrate significant performance gains over state-of-the-art approaches, confirming the method’s effectiveness and robustness.

Abstract

Cycle self-training (CST) breaks the shared classifier assumption of the standard self-training framework, which is effective for unsupervised domain adaptation and exploits unlabeled target data by training with target pseudo-labels. CST introduces a target classifier and employs an inner-outer loop updating strategy, addressing the issue of unreliable pseudo-labels and enabling pseudo-labels to generalize across domains. Despite its success in image classification, extending CST to object detection faces three main challenges. First, the upper bound of CST in object detection is constrained by three types of unreliable pseudo-labels, such as classification error alone, localization error alone, and their combination. Second, since object detection involves detecting multiple target objects, directly applying CST leads to training insta bility. Third, a wider numerical range of regression coordinates leads to exploding losses. To this end, we apply CST to both classification and regression and propose the Dual-Stream Bilevel-Cycle Optimization framework. Specifically, we construct CST upon Mean Teacher to prevent training instability and use extra normalization to map the regression bounding box into a standardized space, effectively addressing exploding losses. Also, we provide a theoretical derivation of the regression bound. Extensive experiments across four cross domain standard scenarios demonstrate that our framework achieves considerable results.

Problem

• Domain Adaptive Object Detection
• Cycle Self-Training
• Unreliable Pseudo-labels
• Training Instability
• Regression Loss Explosion

Innovation

• Domain Adaptive Object Detection
• Cycle Self-Training
• Dual-Stream Bilevel-Cycle Optimization
• Pseudo-Label Refinement
• Regression Normalization