CISO: Co-iteration semi-supervised learning for visual object detection

Abstract Semi-supervised learning offers a solution to the high cost and limited availability of manually labeled samples in supervised learning. In semi-supervised visual object detection, use unlabeled data can significantly enhance performance deep models. this paper, we introduce an end-to-end framework, named CISO (Co-Iteration Semi-Supervised Learning for Object Detection), which integrates knowledge distillation approach collaborative, iterative strategy. To maximize utilization pseudo-label address scarcity due threshold settings, propose mean iteration where all is applied each training iteration. Pseudo-label with confidence extracted based on ever-changing (average intersection over union pseudo-labeled data). This strategy not only ensures accuracy but also optimizes data. Subsequently, apply weak-strong augmentation update model. Lastly, evaluate using Swin Transformer model conduct comprehensive experiments MS-COCO. Our framework showcases impressive results, outperforms state-of-the-art methods by 2.16 mAP 1.54 10% 5% data, respectively.