Addressing Sample Inconsistency for Semisupervised Object Detection in Remote Sensing Images
The emergence of semisupervised object detection (SSOD) techniques has greatly enhanced object detection performance. SSOD leverages a limited amount of labeled data along with a large quantity of unlabeled data. However, there exists a problem of sample inconsistency in remote sensing images, which...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2024-01-01
|
| Series: | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/10463140/ |
| _version_ | 1797238826410704896 |
|---|---|
| author | Yuhao Wang Lifan Yao Gang Meng Xinye Zhang Jiayun Song Haopeng Zhang |
| author_facet | Yuhao Wang Lifan Yao Gang Meng Xinye Zhang Jiayun Song Haopeng Zhang |
| author_sort | Yuhao Wang |
| collection | DOAJ |
| description | The emergence of semisupervised object detection (SSOD) techniques has greatly enhanced object detection performance. SSOD leverages a limited amount of labeled data along with a large quantity of unlabeled data. However, there exists a problem of sample inconsistency in remote sensing images, which manifests in two ways. First, remote sensing images are diverse and complex. Conventional random initialization methods for labeled data are insufficient for training teacher networks to generate high-quality pseudolabels. Finally, remote sensing images typically exhibit a long-tailed distribution, where some categories have a significant number of instances, while others have very few. This distribution poses significant challenges during model training. In this article, we propose the utilization of SSOD networks for remote sensing images characterized by a long-tailed distribution. To address the issue of sample inconsistency between labeled and unlabeled data, we employ a labeled data iterative selection strategy based on the active learning approach. We iteratively filter out high-value samples through the designed selection criteria. The selected samples are labeled and used as data for supervised training. This method filters out valuable labeled data, thereby improving the quality of pseudolabels. Inspired by transfer learning, we decouple the model training into the training of the backbone and the detector. We tackle the problem of sample inconsistency in long-tail distribution data by training the detector using balanced data across categories. Our approach exhibits an approximate 1% improvement over the current state-of-the-art models on both the DOTAv1.0 and DIOR datasets. |
| first_indexed | 2024-04-24T17:41:49Z |
| format | Article |
| id | doaj.art-75e6c9cc125d440399deb7133047ba53 |
| institution | Directory Open Access Journal |
| issn | 2151-1535 |
| language | English |
| last_indexed | 2024-04-24T17:41:49Z |
| publishDate | 2024-01-01 |
| publisher | IEEE |
| record_format | Article |
| series | IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing |
| spelling | doaj.art-75e6c9cc125d440399deb7133047ba532024-03-27T23:00:08ZengIEEEIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing2151-15352024-01-01176933694410.1109/JSTARS.2024.337482010463140Addressing Sample Inconsistency for Semisupervised Object Detection in Remote Sensing ImagesYuhao Wang0https://orcid.org/0009-0005-0764-9771Lifan Yao1https://orcid.org/0009-0007-8734-9879Gang Meng2https://orcid.org/0009-0005-6836-2440Xinye Zhang3https://orcid.org/0000-0003-4264-427XJiayun Song4https://orcid.org/0009-0006-9914-796XHaopeng Zhang5https://orcid.org/0000-0003-1981-8307Department of Aerospace Information Engineering (Image Processing Center), School of Astronautics, Beihang University, Beijing, ChinaQingdao Research Institute of Beihang University, Shandong, ChinaBeijing Institute of Remote Sensing Information, Beijing, ChinaQingdao Research Institute of Beihang University, Shandong, ChinaQingdao Research Institute of Beihang University, Shandong, ChinaDepartment of Aerospace Information Engineering (Image Processing Center), School of Astronautics, Beihang University, Beijing, ChinaThe emergence of semisupervised object detection (SSOD) techniques has greatly enhanced object detection performance. SSOD leverages a limited amount of labeled data along with a large quantity of unlabeled data. However, there exists a problem of sample inconsistency in remote sensing images, which manifests in two ways. First, remote sensing images are diverse and complex. Conventional random initialization methods for labeled data are insufficient for training teacher networks to generate high-quality pseudolabels. Finally, remote sensing images typically exhibit a long-tailed distribution, where some categories have a significant number of instances, while others have very few. This distribution poses significant challenges during model training. In this article, we propose the utilization of SSOD networks for remote sensing images characterized by a long-tailed distribution. To address the issue of sample inconsistency between labeled and unlabeled data, we employ a labeled data iterative selection strategy based on the active learning approach. We iteratively filter out high-value samples through the designed selection criteria. The selected samples are labeled and used as data for supervised training. This method filters out valuable labeled data, thereby improving the quality of pseudolabels. Inspired by transfer learning, we decouple the model training into the training of the backbone and the detector. We tackle the problem of sample inconsistency in long-tail distribution data by training the detector using balanced data across categories. Our approach exhibits an approximate 1% improvement over the current state-of-the-art models on both the DOTAv1.0 and DIOR datasets.https://ieeexplore.ieee.org/document/10463140/Active learninglong-tailed distributionremote sensingsemisupervised object detection (SSOD) |
| spellingShingle | Yuhao Wang Lifan Yao Gang Meng Xinye Zhang Jiayun Song Haopeng Zhang Addressing Sample Inconsistency for Semisupervised Object Detection in Remote Sensing Images IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing Active learning long-tailed distribution remote sensing semisupervised object detection (SSOD) |
| title | Addressing Sample Inconsistency for Semisupervised Object Detection in Remote Sensing Images |
| title_full | Addressing Sample Inconsistency for Semisupervised Object Detection in Remote Sensing Images |
| title_fullStr | Addressing Sample Inconsistency for Semisupervised Object Detection in Remote Sensing Images |
| title_full_unstemmed | Addressing Sample Inconsistency for Semisupervised Object Detection in Remote Sensing Images |
| title_short | Addressing Sample Inconsistency for Semisupervised Object Detection in Remote Sensing Images |
| title_sort | addressing sample inconsistency for semisupervised object detection in remote sensing images |
| topic | Active learning long-tailed distribution remote sensing semisupervised object detection (SSOD) |
| url | https://ieeexplore.ieee.org/document/10463140/ |
| work_keys_str_mv | AT yuhaowang addressingsampleinconsistencyforsemisupervisedobjectdetectioninremotesensingimages AT lifanyao addressingsampleinconsistencyforsemisupervisedobjectdetectioninremotesensingimages AT gangmeng addressingsampleinconsistencyforsemisupervisedobjectdetectioninremotesensingimages AT xinyezhang addressingsampleinconsistencyforsemisupervisedobjectdetectioninremotesensingimages AT jiayunsong addressingsampleinconsistencyforsemisupervisedobjectdetectioninremotesensingimages AT haopengzhang addressingsampleinconsistencyforsemisupervisedobjectdetectioninremotesensingimages |