关键词: 半监督学习, 目标检测, 课程学习, 训练策略, 难度测量器, 训练调度器

Abstract: Abstract: In order to enhance the quality of pseudo labels, address the issue of confirmation bias in semi-supervised object detection, and tackle the challenge of ignoring complexities in unlabeled data leading to erroneous pseudo labels in existing algorithms, a semi-supervised object detection framework guided by curriculum learning is proposed. The framework consists of two modules: the ICSD (Iou-Confidence-Standard-Deviation) difficulty measurer and the BP (Batch-Package) trainingscheduler. The ICSD difficulty measurer comprehensively considers information such as IOU, confidence, and class labels between pseudo-bounding boxes，and introduces the C_IOU (Checkpoint_IOU) method to evaluate the reliability of unlabeled data. The BP training scheduler has designed two efficient scheduling strategies, starting from the perspective of Batch and Package respectively, giving priority to unlabeled data with high reliability indicators to achieve full utilization of the entire unlabeled data set in the form of course learning. Extensive comparative experiments on the PASCAL VOC and MS-COCO datasets demonstrate that this framework applies to existing semi-supervised object detection algorithms and exhibits significant improvements in detection accuracy and stability.

Key words: semi-supervised learning, object detection, curriculum learning, training strategy, difficulty measurer, training scheduler