期刊
2022 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2022)
卷 -, 期 -, 页码 14492-14501出版社
IEEE COMPUTER SOC
DOI: 10.1109/CVPR52688.2022.01411
关键词
-
资金
- NRF [2021R1A2C3006659]
- IITP - Korea government (MSIT) [2021-0-01343]
Many recent semi-supervised learning studies utilize a teacher-student architecture to train the student network with supervisory signals from the teacher. Data augmentation plays a vital role in SSL frameworks, especially when extending SSL to semi-supervised object detection. To address this, researchers have introduced a simple yet effective data augmentation method called Mix/UnMix (MUM) for the SSOD framework, improving its performance over baseline methods.
Many recent semi-supervised learning (SSL) studies build teacher-student architecture and train the student network by the generated supervisory signal from the teacher. Data augmentation strategy plays a significant role in the SSL framework since it is hard to create a weak-strong augmented input pair without losing label information. Especially when extending SSL to semi-supervised object detection (SSOD), many strong augmentation methodologies related to image geometry and interpolation-regularization are hard to utilize since they possibly hurt the location information of the bounding box in the object detection task. To address this, we introduce a simple yet effective data augmentation method, Mix/UnMix (MUM), which unmixes feature tiles for the mixed image tiles for the SSOD framework. Our proposed method makes mixed input image tiles and reconstructs them in the feature space. Thus, MUM can enjoy the interpolation-regularization effect from non-interpolated pseudo-labels and successfully generate a meaningful weak-strong pair. Furthermore, MUM can be easily equipped on top of various SSOD methods. Extensive experiments on MS-COCO and PASCAL VOC datasets demonstrate the superiority of MUM by consistently improving the mAP performance over the baseline in all the tested SSOD benchmark protocols. The code is released at https://github.com/JongMokKim/mix-unmix.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据