论文标题

一致的老师:在半监督对象检测中减少不一致的伪靶

Consistent-Teacher: Towards Reducing Inconsistent Pseudo-targets in Semi-supervised Object Detection

论文作者

Wang, Xinjiang, Yang, Xingyi, Zhang, Shilong, Li, Yijiang, Feng, Litong, Fang, Shijie, Lyu, Chengqi, Chen, Kai, Zhang, Wayne

论文摘要

在这项研究中,我们深入研究了半监督物体检测(SSOD)中伪靶标的不一致。我们的核心观察是,振荡伪靶标破坏了精确检测器的训练。它向学生的培训注入噪音,导致严重的过度拟合问题。因此,我们提出了一个系统的解决方案,称为一致的教师,以减少不一致。首先,自适应锚分配〜(ASA)代替了基于静态的策略,该策略使学生网络能够抵抗噪音伪装的框。然后,我们通过设计3D特征对齐模块〜(FAM-3D)来校准子任务预测。它允许每个分类功能在任意尺度和位置的回归任务中适应最佳特征向量。最后,高斯混合模型(GMM)动态修改了伪基盒的得分阈值,该模型在早期稳定了地面真相的数量,并在训练过程中补救了不可靠的监督信号。一致的教师在各种SSOD评估上都提供了良好的结果。只有10%的带注释的MS-Coco数据,它可以使用Resnet-50骨干实现40.0 MAP,该数据使用伪标签超过了先前的基准,该数据超过了先前的基准。当对完全注释的MS-Coco培训并使用其他未标记的数据进行培训时,性能将进一步增加到47.7 MAP。我们的代码可在\ url {https://github.com/adamdad/consistentteacher}上找到。

In this study, we dive deep into the inconsistency of pseudo targets in semi-supervised object detection (SSOD). Our core observation is that the oscillating pseudo-targets undermine the training of an accurate detector. It injects noise into the student's training, leading to severe overfitting problems. Therefore, we propose a systematic solution, termed ConsistentTeacher, to reduce the inconsistency. First, adaptive anchor assignment~(ASA) substitutes the static IoU-based strategy, which enables the student network to be resistant to noisy pseudo-bounding boxes. Then we calibrate the subtask predictions by designing a 3D feature alignment module~(FAM-3D). It allows each classification feature to adaptively query the optimal feature vector for the regression task at arbitrary scales and locations. Lastly, a Gaussian Mixture Model (GMM) dynamically revises the score threshold of pseudo-bboxes, which stabilizes the number of ground truths at an early stage and remedies the unreliable supervision signal during training. ConsistentTeacher provides strong results on a large range of SSOD evaluations. It achieves 40.0 mAP with ResNet-50 backbone given only 10% of annotated MS-COCO data, which surpasses previous baselines using pseudo labels by around 3 mAP. When trained on fully annotated MS-COCO with additional unlabeled data, the performance further increases to 47.7 mAP. Our code is available at \url{https://github.com/Adamdad/ConsistentTeacher}.

扫码加入交流群

加入微信交流群

微信交流群二维码

扫码加入学术交流群,获取更多资源