Semi-Supervised Object Detection with Adaptive Class-Rebalancing Self-Training

TL;DR

This paper introduces an adaptive class-rebalancing self-training method with a novel memory module, CropBank, to address class imbalance and noisy pseudo-labels in semi-supervised object detection, significantly improving performance on benchmarks.

Contribution

It proposes ACRST with CropBank to effectively rebalance training data and a two-stage filtering for accurate pseudo-labels, advancing semi-supervised object detection techniques.

Findings

Achieves 17.02 mAP improvement on MS-COCO with 1% labeled data.

Outperforms state-of-the-art semi-supervised methods by 5.32 mAP.

Demonstrates effectiveness of class rebalancing and label filtering in SSOD.

Abstract

This study delves into semi-supervised object detection (SSOD) to improve detector performance with additional unlabeled data. State-of-the-art SSOD performance has been achieved recently by self-training, in which training supervision consists of ground truths and pseudo-labels. In current studies, we observe that class imbalance in SSOD severely impedes the effectiveness of self-training. To address the class imbalance, we propose adaptive class-rebalancing self-training (ACRST) with a novel memory module called CropBank. ACRST adaptively rebalances the training data with foreground instances extracted from the CropBank, thereby alleviating the class imbalance. Owing to the high complexity of detection tasks, we observe that both self-training and data-rebalancing suffer from noisy pseudo-labels in SSOD. Therefore, we propose a novel two-stage filtering algorithm to generate accurate pseudo-labels. Our method achieves satisfactory improvements on MS-COCO and VOC benchmarks. When using only 1\% labeled data in MS-COCO, our method achieves 17.02 mAP improvement over supervised baselines, and 5.32 mAP improvement compared with state-of-the-art methods.