Training-based Model Refinement and Representation Disagreement for Semi-Supervised Object Detection

TL;DR

This paper introduces a training-based model refinement stage and a representation disagreement strategy to enhance semi-supervised object detection, addressing issues of model overfitting, consensus, and pseudo-label noise, leading to improved performance.

Contribution

It proposes novel TMR and RD strategies that can be integrated into existing SSOD methods to improve model refinement and encourage diversity in teacher-student models.

Findings

Outperforms state-of-the-art SSOD methods by 2-3.36 mAP on COCO and Pascal datasets.

Effectively refines pseudo-labels and model weights using TMR.

Encourages model divergence with RD to explore more unlabeled data patterns.

Abstract

Semi-supervised object detection (SSOD) aims to improve the performance and generalization of existing object detectors by utilizing limited labeled data and extensive unlabeled data. Despite many advances, recent SSOD methods are still challenged by inadequate model refinement using the classical exponential moving average (EMA) strategy, the consensus of Teacher-Student models in the latter stages of training (i.e., losing their distinctiveness), and noisy/misleading pseudo-labels. This paper proposes a novel training-based model refinement (TMR) stage and a simple yet effective representation disagreement (RD) strategy to address the limitations of classical EMA and the consensus problem. The TMR stage of Teacher-Student models optimizes the lightweight scaling operation to refine the model's weights and prevent overfitting or forgetting learned patterns from unlabeled data. Meanwhile, the RD strategy helps keep these models diverged to encourage the student model to explore additional patterns in unlabeled data. Our approach can be integrated into established SSOD methods and is empirically validated using two baseline methods, with and without cascade regression, to generate more reliable pseudo-labels. Extensive experiments demonstrate the superior performance of our approach over state-of-the-art SSOD methods. Specifically, the proposed approach outperforms the baseline Unbiased-Teacher-v2 (& Unbiased-Teacher-v1) method by an average mAP margin of 2.23, 2.1, and 3.36 (& 2.07, 1.9, and 3.27) on COCO-standard, COCO-additional, and Pascal VOC datasets, respectively.