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

While self-training achieves state-of-the-art results in semi-supervised object detection (SSOD), it severely suffers from foreground-background and foreground-foreground imbalances SSOD. In this paper, we propose an Adaptive Class-Rebalancing Self-Training (ACRST) with a novel memory module called CropBank to alleviate these generate unbiased pseudo-labels. Besides, observe that both data-rebalancing procedures suffer noisy pseudo-labels Therefore, contribute simple yet effective two-stage pseudo-label filtering scheme obtain accurate supervision. Our method competitive performance on MS-COCO VOC benchmarks. When using only 1% labeled data of MS-COCO, our 17.02 mAP improvement over the supervised 5.32 gains compared state-of-the-arts.