【摘要】 半监督目标检测方法的关键是为无标签数据中的目标分配伪标签。为了确保伪标签的质量，半监督目标检测方法通常利用置信度阈值过滤不可靠伪标签，这会导致大量的伪标签因为置信度低而被滤除。本文提出改进后的半监督学习方法使用对比学习来重利用大量置信度低的不可靠伪标签，提升半监督目标检测方法的性能。具体来说，根据预测置信度将无标签数据的伪标签分为可靠与不可靠伪标签。除了利用可靠伪标签，还利用不可靠伪标签作为对比学习中的负样本训练模型。为了平衡类别间不可靠伪标签的数量，设计一个记忆模块用于保存训练过程中不同批次的不可靠伪标签。实验结果表明，在COCO数据集上，对训练数据进行1%、5%和10%的标注情况下，改进后的半监督学习方法的平均准确率达到13.6%、23.0%和27.5%，优于已有半监督学习方法；在COCO-additional数据集上，改进后的半监督学习方法的平均准确率达到44.7%，相较于监督学习，性能提高4.5个百分点。更多还原

【Abstract】 The key to semi-supervised object detection methods is to assign pseudo labels to the targets of unlabeled data. Toguarantee the quality of pseudo-labels,the semi-supervised object detection methods usually use a confidence threshold to filterlow-quality pseudo-labels,which will cause most pseudo-labels to be removed due to their low confidence. Contrastive learningis used to reuse most of low-confidence unreliable pseudo labels for boosting the performance of semi-supervised object detectionmethod. Specifically,the pseudo-labels are divided into reliable and unreliable ones according to the prediction confidence. Be-sides the reliable pseudo-labels,the unreliable pseudo-labels are exploited as negative samples for model training of contrastlearning. To balance the number of unreliable pseudo-labels between different classes,a memory module is designed to store theunreliable pseudo-labels of different batches in the training process. The experimental results show that the mAP of the improvedsemi-supervised method on COCO data set is 13.6%,23.0%,and 27.5% with the labeling ratio of 1%,5%,and 10%,which isbetter than the existing semi-supervised learning methods. On the COCO-additional data set,the mAP of the improved semi-supervised method reaches 44.7%,which is 4.5 percentage points higher than supervised learning.更多还原