Weakly Supervised Deep Learning for Segmentation of Remote Sensing Imagery
Accurate automated segmentation of remote sensing data could benefit applications from land cover mapping and agricultural monitoring to urban development surveyal and disaster damage assessment. While convolutional neural networks (CNNs) achieve state-of-the-art accuracy when segmenting natural ima...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-01-01
|
| Series: | Remote Sensing |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2072-4292/12/2/207 |
| _version_ | 1818321617663557632 |
|---|---|
| author | Sherrie Wang William Chen Sang Michael Xie George Azzari David B. Lobell |
| author_facet | Sherrie Wang William Chen Sang Michael Xie George Azzari David B. Lobell |
| author_sort | Sherrie Wang |
| collection | DOAJ |
| description | Accurate automated segmentation of remote sensing data could benefit applications from land cover mapping and agricultural monitoring to urban development surveyal and disaster damage assessment. While convolutional neural networks (CNNs) achieve state-of-the-art accuracy when segmenting natural images with huge labeled datasets, their successful translation to remote sensing tasks has been limited by low quantities of ground truth labels, especially fully segmented ones, in the remote sensing domain. In this work, we perform cropland segmentation using two types of labels commonly found in remote sensing datasets that can be considered sources of “weak supervision”: (1) labels comprised of single geotagged points and (2) image-level labels. We demonstrate that (1) a U-Net trained on a single labeled pixel per image and (2) a U-Net image classifier transferred to segmentation can outperform pixel-level algorithms such as logistic regression, support vector machine, and random forest. While the high performance of neural networks is well-established for large datasets, our experiments indicate that U-Nets trained on weak labels outperform baseline methods with as few as 100 labels. Neural networks, therefore, can combine superior classification performance with efficient label usage, and allow pixel-level labels to be obtained from image labels. |
| first_indexed | 2024-12-13T10:43:45Z |
| format | Article |
| id | doaj.art-3b248a42288a4f9db31eacba8b958274 |
| institution | Directory Open Access Journal |
| issn | 2072-4292 |
| language | English |
| last_indexed | 2024-12-13T10:43:45Z |
| publishDate | 2020-01-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Remote Sensing |
| spelling | doaj.art-3b248a42288a4f9db31eacba8b9582742022-12-21T23:50:19ZengMDPI AGRemote Sensing2072-42922020-01-0112220710.3390/rs12020207rs12020207Weakly Supervised Deep Learning for Segmentation of Remote Sensing ImagerySherrie Wang0William Chen1Sang Michael Xie2George Azzari3David B. Lobell4Institute for Computational and Mathematical Engineering, Stanford University, Stanford, CA 94305, USADepartment of Computer Science, Stanford University, Stanford, CA 94305, USADepartment of Computer Science, Stanford University, Stanford, CA 94305, USADepartment of Earth System Science, Stanford University, Stanford, CA 94305, USADepartment of Earth System Science, Stanford University, Stanford, CA 94305, USAAccurate automated segmentation of remote sensing data could benefit applications from land cover mapping and agricultural monitoring to urban development surveyal and disaster damage assessment. While convolutional neural networks (CNNs) achieve state-of-the-art accuracy when segmenting natural images with huge labeled datasets, their successful translation to remote sensing tasks has been limited by low quantities of ground truth labels, especially fully segmented ones, in the remote sensing domain. In this work, we perform cropland segmentation using two types of labels commonly found in remote sensing datasets that can be considered sources of “weak supervision”: (1) labels comprised of single geotagged points and (2) image-level labels. We demonstrate that (1) a U-Net trained on a single labeled pixel per image and (2) a U-Net image classifier transferred to segmentation can outperform pixel-level algorithms such as logistic regression, support vector machine, and random forest. While the high performance of neural networks is well-established for large datasets, our experiments indicate that U-Nets trained on weak labels outperform baseline methods with as few as 100 labels. Neural networks, therefore, can combine superior classification performance with efficient label usage, and allow pixel-level labels to be obtained from image labels.https://www.mdpi.com/2072-4292/12/2/207deep learningimage segmentationweak supervisionagriculturelandsatland cover classification |
| spellingShingle | Sherrie Wang William Chen Sang Michael Xie George Azzari David B. Lobell Weakly Supervised Deep Learning for Segmentation of Remote Sensing Imagery Remote Sensing deep learning image segmentation weak supervision agriculture landsat land cover classification |
| title | Weakly Supervised Deep Learning for Segmentation of Remote Sensing Imagery |
| title_full | Weakly Supervised Deep Learning for Segmentation of Remote Sensing Imagery |
| title_fullStr | Weakly Supervised Deep Learning for Segmentation of Remote Sensing Imagery |
| title_full_unstemmed | Weakly Supervised Deep Learning for Segmentation of Remote Sensing Imagery |
| title_short | Weakly Supervised Deep Learning for Segmentation of Remote Sensing Imagery |
| title_sort | weakly supervised deep learning for segmentation of remote sensing imagery |
| topic | deep learning image segmentation weak supervision agriculture landsat land cover classification |
| url | https://www.mdpi.com/2072-4292/12/2/207 |
| work_keys_str_mv | AT sherriewang weaklysuperviseddeeplearningforsegmentationofremotesensingimagery AT williamchen weaklysuperviseddeeplearningforsegmentationofremotesensingimagery AT sangmichaelxie weaklysuperviseddeeplearningforsegmentationofremotesensingimagery AT georgeazzari weaklysuperviseddeeplearningforsegmentationofremotesensingimagery AT davidblobell weaklysuperviseddeeplearningforsegmentationofremotesensingimagery |