Enhanced Feature Pyramid Vision Transformer for Semantic Segmentation on Thailand Landsat-8 Corpus
Semantic segmentation on Landsat-8 data is crucial in the integration of diverse data, allowing researchers to achieve more productivity and lower expenses. This research aimed to improve the versatile backbone for dense prediction without convolutions—namely, using the pyramid vision transformer (P...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-05-01
|
| Series: | Information |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2078-2489/13/5/259 |
| _version_ | 1797499037837950976 |
|---|---|
| author | Kritchayan Intarat Preesan Rakwatin Teerapong Panboonyuen |
| author_facet | Kritchayan Intarat Preesan Rakwatin Teerapong Panboonyuen |
| author_sort | Kritchayan Intarat |
| collection | DOAJ |
| description | Semantic segmentation on Landsat-8 data is crucial in the integration of diverse data, allowing researchers to achieve more productivity and lower expenses. This research aimed to improve the versatile backbone for dense prediction without convolutions—namely, using the pyramid vision transformer (PRM-VS-TM) to incorporate attention mechanisms across various feature maps. Furthermore, the PRM-VS-TM constructs an end-to-end object detection system without convolutions and uses handcrafted components, such as dense anchors and non-maximum suspension (NMS). The present study was conducted on a private dataset, i.e., the Thailand Landsat-8 challenge. There are three baselines: DeepLab, Swin Transformer (Swin TF), and PRM-VS-TM. Results indicate that the proposed model significantly outperforms all current baselines on the Thailand Landsat-8 corpus, providing F1-scores greater than 80% in almost all categories. Finally, we demonstrate that our model, without utilizing pre-trained settings or any further post-processing, can outperform current state-of-the-art (SOTA) methods for both agriculture and forest classes. |
| first_indexed | 2024-03-10T03:41:46Z |
| format | Article |
| id | doaj.art-ae54d6d562eb47a9a6c4030d29d602a4 |
| institution | Directory Open Access Journal |
| issn | 2078-2489 |
| language | English |
| last_indexed | 2024-03-10T03:41:46Z |
| publishDate | 2022-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Information |
| spelling | doaj.art-ae54d6d562eb47a9a6c4030d29d602a42023-11-23T11:30:28ZengMDPI AGInformation2078-24892022-05-0113525910.3390/info13050259Enhanced Feature Pyramid Vision Transformer for Semantic Segmentation on Thailand Landsat-8 CorpusKritchayan Intarat0Preesan Rakwatin1Teerapong Panboonyuen2Department of Geography, Faculty of Liberal Arts, Thammasat University, 99 Moo 18, Phahonyothin Rd., Khlong Luang, Khlong Nueng, Pathum Thani 12121, ThailandDigital Economy Promotion Agency, 80 Soi Ladpro 4, Ladprao Rd., Chom Phon, Chatuchak, Bangkok 10900, ThailandDepartment of Geography, Faculty of Liberal Arts, Thammasat University, 99 Moo 18, Phahonyothin Rd., Khlong Luang, Khlong Nueng, Pathum Thani 12121, ThailandSemantic segmentation on Landsat-8 data is crucial in the integration of diverse data, allowing researchers to achieve more productivity and lower expenses. This research aimed to improve the versatile backbone for dense prediction without convolutions—namely, using the pyramid vision transformer (PRM-VS-TM) to incorporate attention mechanisms across various feature maps. Furthermore, the PRM-VS-TM constructs an end-to-end object detection system without convolutions and uses handcrafted components, such as dense anchors and non-maximum suspension (NMS). The present study was conducted on a private dataset, i.e., the Thailand Landsat-8 challenge. There are three baselines: DeepLab, Swin Transformer (Swin TF), and PRM-VS-TM. Results indicate that the proposed model significantly outperforms all current baselines on the Thailand Landsat-8 corpus, providing F1-scores greater than 80% in almost all categories. Finally, we demonstrate that our model, without utilizing pre-trained settings or any further post-processing, can outperform current state-of-the-art (SOTA) methods for both agriculture and forest classes.https://www.mdpi.com/2078-2489/13/5/259deep learningpyramid vision transformerLandsat-8satellite imageattention |
| spellingShingle | Kritchayan Intarat Preesan Rakwatin Teerapong Panboonyuen Enhanced Feature Pyramid Vision Transformer for Semantic Segmentation on Thailand Landsat-8 Corpus Information deep learning pyramid vision transformer Landsat-8 satellite image attention |
| title | Enhanced Feature Pyramid Vision Transformer for Semantic Segmentation on Thailand Landsat-8 Corpus |
| title_full | Enhanced Feature Pyramid Vision Transformer for Semantic Segmentation on Thailand Landsat-8 Corpus |
| title_fullStr | Enhanced Feature Pyramid Vision Transformer for Semantic Segmentation on Thailand Landsat-8 Corpus |
| title_full_unstemmed | Enhanced Feature Pyramid Vision Transformer for Semantic Segmentation on Thailand Landsat-8 Corpus |
| title_short | Enhanced Feature Pyramid Vision Transformer for Semantic Segmentation on Thailand Landsat-8 Corpus |
| title_sort | enhanced feature pyramid vision transformer for semantic segmentation on thailand landsat 8 corpus |
| topic | deep learning pyramid vision transformer Landsat-8 satellite image attention |
| url | https://www.mdpi.com/2078-2489/13/5/259 |
| work_keys_str_mv | AT kritchayanintarat enhancedfeaturepyramidvisiontransformerforsemanticsegmentationonthailandlandsat8corpus AT preesanrakwatin enhancedfeaturepyramidvisiontransformerforsemanticsegmentationonthailandlandsat8corpus AT teerapongpanboonyuen enhancedfeaturepyramidvisiontransformerforsemanticsegmentationonthailandlandsat8corpus |