{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,11,5]],"date-time":"2025-11-05T07:27:16Z","timestamp":1762327636367,"version":"build-2065373602"},"reference-count":41,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2025,11,5]],"date-time":"2025-11-05T00:00:00Z","timestamp":1762300800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Comput. Sci."],"abstract":"<jats:p>Railway image classification (RIC) represents a critical application in railway infrastructure monitoring, involving the analysis of hyperspectral datasets with complex spatial-spectral relationships unique to railway environments. Nevertheless, Transformer-based methodologies for RIC face obstacles pertaining to the extraction of local features and the efficiency of training processes. To address these challenges, we introduce the Pure Transformer Network (PTN), an entirely Transformer-centric framework tailored for the effective execution of RIC tasks. Our approach improves the amalgamation of local and global data within railway images by utilizing a Patch Embedding Transformer (PET) module that employs an \u201cunfold + attention + fold\u201d mechanism in conjunction with a Transformer module that incorporates relative attention. The PET module harnesses attention mechanisms to replicate convolutional operations, enabling adaptive receptive fields for varying spatial patterns in railway infrastructure, thus circumventing the constraints imposed by fixed convolutional kernels. Additionally, we propose a Memory Efficient Algorithm that achieves 35% training time reduction while preserving accuracy. Thorough assessments conducted on four hyperspectral railway image datasets validate the PTN's exceptional performance, demonstrating superior accuracy compared to existing CNN- and Transformer-based baselines.<\/jats:p>","DOI":"10.3389\/fcomp.2025.1658556","type":"journal-article","created":{"date-parts":[[2025,11,5]],"date-time":"2025-11-05T06:29:07Z","timestamp":1762324147000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Optimized encoder-based transformers for improved local and global integration in railway image classification"],"prefix":"10.3389","volume":"7","author":[{"given":"Lilan","family":"Li","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Xuemei","family":"Zhan","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"TianTian","family":"Wu","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Hua","family":"Ma","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1965","published-online":{"date-parts":[[2025,11,5]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"4301","DOI":"10.1016\/j.rse.2008.07.016","article-title":"The role of environmental context in mapping invasive plants with hyperspectral image data","volume":"112","author":"Andrew","year":"2008","journal-title":"Remote Sens. Environ"},{"key":"B2","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/TGRS.2022.3196661","article-title":"Hyperspectral image classification based on multibranch attention transformer networks","volume":"60","author":"Bai","year":"2022","journal-title":"IEEE Trans. Geosci. Remote Sens"},{"key":"B3","first-page":"357","article-title":"\u201cCrossvit: Cross-attention multi-scale vision transformer for image classification,\u201d","volume-title":"Proceedings of the IEEE\/CVF International Conference on Computer Vision","author":"Chen","year":"2021"},{"key":"B4","doi-asserted-by":"publisher","first-page":"6232","DOI":"10.1109\/TGRS.2016.2584107","article-title":"Deep feature extraction and classification of hyperspectral images based on convolutional neural networks","volume":"54","author":"Chen","year":"2016","journal-title":"IEEE Trans. Geosci. Remote Sens"},{"key":"B5","doi-asserted-by":"publisher","first-page":"2094","DOI":"10.1109\/JSTARS.2014.2329330","article-title":"Deep learning-based classification of hyperspectral data","volume":"7","author":"Chen","year":"2014","journal-title":"IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens"},{"key":"B6","doi-asserted-by":"publisher","first-page":"2381","DOI":"10.1109\/JSTARS.2015.2388577","article-title":"Spectral-spatial classification of hyperspectral data based on deep belief network","volume":"8","author":"Chen","year":"2015","journal-title":"IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens"},{"key":"B7","first-page":"3965","article-title":"\u201cCoatNet: Marrying convolution and attention for all data sizes,\u201d","author":"Dai","year":"2021","journal-title":"Advances in Neural Information Processing Systems"},{"key":"B8","first-page":"16344","article-title":"Flashattention: fast and memory-efficient exact attention with IO-awareness","volume":"35","author":"Dao","year":"2022","journal-title":"Adv. Neural Inform. Process. Syst."},{"key":"B9","first-page":"1","article-title":"Switch transformers: scaling to trillion parameter models with simple and efficient sparsity","volume":"23","author":"Fedus","year":"2022","journal-title":"J. Mach. Learn. Res."},{"key":"B10","doi-asserted-by":"publisher","first-page":"165","DOI":"10.1109\/TGRS.2019.2934760","article-title":"Hsi-bert: Hyperspectral image classification using the bidirectional encoder representation from transformers","volume":"58","author":"He","year":"2020","journal-title":"IEEE Trans. Geosci. Remote Sens"},{"key":"B11","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/TGRS.2021.3130716","article-title":"Spectralformer: Rethinking hyperspectral image classification with transformers","volume":"60","author":"Hong","year":"2022","journal-title":"IEEE Trans. Geosci. Remote Sens"},{"key":"B12","doi-asserted-by":"crossref","first-page":"7132","DOI":"10.1109\/CVPR.2018.00745","article-title":"\u201cSqueeze-and-excitation networks,\u201d","volume-title":"2018 IEEE\/CVF Conference on Computer Vision and Pattern Recognition","author":"Hu","year":"2018"},{"key":"B13","doi-asserted-by":"publisher","first-page":"258619","DOI":"10.1155\/2015\/258619","article-title":"Deep convolutional neural networks for hyperspectral image classification","volume":"2015","author":"Hu","year":"2015","journal-title":"J. Sens"},{"key":"B14","author":"Huang","year":"2018","journal-title":"Music Transformer"},{"key":"B15","doi-asserted-by":"publisher","first-page":"1366","DOI":"10.3390\/rs10091366","article-title":"Integration of terrestrial and drone-borne hyperspectral and photogrammetric sensing methods for exploration mapping and mining monitoring","volume":"10","author":"Kirsch","year":"2018","journal-title":"Remote Sens"},{"key":"B16","doi-asserted-by":"publisher","first-page":"1330","DOI":"10.3390\/rs9121330","article-title":"Spectral-spatial classification of hyperspectral imagery with 3D convolutional neural network","volume":"9","author":"Li","year":"2017","journal-title":"Remote Sens"},{"key":"B17","first-page":"10012","article-title":"\u201cSwin transformer: Hierarchical vision transformer using shifted windows,\u201d","volume-title":"Proceedings of the IEEE\/CVF International Conference on Computer Vision","author":"Liu","year":"2021"},{"key":"B18","doi-asserted-by":"publisher","first-page":"4099","DOI":"10.1109\/TGRS.2010.2055876","article-title":"Local manifold learning-based k-nearest-neighbor for hyperspectral image classification","volume":"48","author":"Ma","year":"2010","journal-title":"IEEE Trans. Geosci. Remote Sens"},{"key":"B19","doi-asserted-by":"publisher","first-page":"17","DOI":"10.1016\/j.jspr.2015.01.006","article-title":"Hyperspectral imaging to classify and monitor quality of agricultural materials","volume":"61","author":"Mahesh","year":"2015","journal-title":"J. Stored Prod. Res"},{"key":"B20","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/TGRS.2022.3207933","article-title":"Hyperspectral image classification using group-aware hierarchical transformer","volume":"60","author":"Mei","year":"2022","journal-title":"IEEE Trans. Geosci. Remote Sens"},{"key":"B21","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1710.03740","article-title":"Mixed precision training","author":"Micikevicius","year":"2017","journal-title":"arXiv"},{"key":"B22","author":"Platt","year":"1998","journal-title":"Sequential Minimal Optimization: A Fast Algorithm for Training Support Vector Machines"},{"key":"B23","doi-asserted-by":"publisher","first-page":"1297","DOI":"10.1080\/10408398.2022.2121805","article-title":"Recent advances in muscle food safety evaluation: Hyperspectral imaging analyses and applications","volume":"63","author":"Pu","year":"2023","journal-title":"Crit. Rev. Food Sci. Nutr"},{"key":"B24","doi-asserted-by":"publisher","first-page":"2216","DOI":"10.3390\/rs13112216","article-title":"Improved transformer net for hyperspectral image classification","volume":"13","author":"Qing","year":"2021","journal-title":"Remote Sens"},{"key":"B25","doi-asserted-by":"publisher","first-page":"277","DOI":"10.1109\/LGRS.2019.2918719","article-title":"HybridSN: Exploring 3-D-2-D CNN feature hierarchy for hyperspectral image classification","volume":"17","author":"Roy","year":"2020","journal-title":"IEEE Geosci. Remote Sens. Lett"},{"key":"B26","doi-asserted-by":"publisher","first-page":"106325","DOI":"10.1016\/j.compag.2021.106325","article-title":"Extraction of spatial-spectral homogeneous patches and fractional abundances for field-scale agriculture monitoring using airborne hyperspectral images","volume":"188","author":"Sahadevan","year":"2021","journal-title":"Comput. Electron. Agric"},{"key":"B27","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/TGRS.2022.3144158","article-title":"Spectral-spatial feature tokenization transformer for hyperspectral image classification","volume":"60","author":"Sun","year":"2022","journal-title":"IEEE Trans. Geosci. Remote Sens"},{"key":"B28","first-page":"10347","article-title":"\u201cTraining data-efficient image transformers & distillation through attention,\u201d","volume-title":"International Conference on Machine Learning","author":"Touvron","year":"2021"},{"key":"B29","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1706.03762","article-title":"Attention is all you need","author":"Vaswani","year":"2017","journal-title":"arXiv"},{"key":"B30","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/TIM.2023.3293548","article-title":"Deep margin cosine autoencoder-based medical hyperspectral image classification for tumor diagnosis","volume":"72","author":"Wang","year":"2023","journal-title":"IEEE Trans. Instrum. Meas"},{"key":"B31","doi-asserted-by":"publisher","first-page":"415","DOI":"10.1007\/s41095-022-0274-8","article-title":"Pvt v2: Improved baselines with pyramid vision transformer","volume":"8","author":"Wang","year":"2022","journal-title":"Comput. Vis. Media"},{"key":"B32","first-page":"22","article-title":"\u201cCvT: Introducing convolutions to vision transformers,\u201d","volume-title":"Proceedings of the IEEE\/CVF International Conference on Computer Vision","author":"Wu","year":"2021"},{"key":"B33","doi-asserted-by":"publisher","first-page":"202","DOI":"10.1109\/TGRS.2017.2744662","article-title":"Random forest ensembles and extended multiextinction profiles for hyperspectral image classification","volume":"56","author":"Xia","year":"2018","journal-title":"IEEE Trans. Geosci. Remote Sens"},{"key":"B34","doi-asserted-by":"publisher","first-page":"1101","DOI":"10.1016\/S0308-521X(02)00051-3","article-title":"Application of decision tree technology for image classification using remote sensing data","volume":"76","author":"Yang","year":"2003","journal-title":"Agric. Syst"},{"key":"B35","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/TGRS.2022.3171551","article-title":"Hyperspectral image transformer classification networks","volume":"60","author":"Yang","year":"2022","journal-title":"IEEE Trans. Geosci. Remote Sens"},{"key":"B36","doi-asserted-by":"publisher","first-page":"5408","DOI":"10.1109\/TGRS.2018.2815613","article-title":"Hyperspectral image classification with deep learning models","volume":"56","author":"Yang","year":"2018","journal-title":"IEEE Trans. Geosci. Remote Sens"},{"key":"B37","doi-asserted-by":"crossref","first-page":"149","DOI":"10.1007\/978-981-13-6504-1_20","article-title":"\u201cDeep 2D convolutional neural network with deconvolution layer for hyperspectral image classification,\u201d","volume-title":"Communications, Signal Processing, and Systems, Lecture Notes in Electrical Engineering","author":"Yu","year":"2020"},{"key":"B38","doi-asserted-by":"crossref","first-page":"538","DOI":"10.1109\/ICCV48922.2021.00060","article-title":"\u201cTokens-to-token vit: Training vision transformers from scratch on imagenet,\u201d","volume-title":"2021 IEEE\/CVF International Conference on Computer Vision (ICCV)","author":"Yuan","year":"2021"},{"key":"B39","doi-asserted-by":"publisher","DOI":"10.24963\/ijcai.2021\/472","article-title":"\u201cAutomatic mixed-precision quantization search of bert,\u201d","author":"Zhao","year":"2021","journal-title":"Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence"},{"key":"B40","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/TGRS.2021.3115699","article-title":"Spectral-spatial transformer network for hyperspectral image classification: a factorized architecture search framework","volume":"60","author":"Zhong","year":"2022","journal-title":"IEEE Trans. Geosci. Remote Sens"},{"key":"B41","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1109\/TGRS.2022.3196771","article-title":"Lessformer: Local-enhanced spectral-spatial transformer for hyperspectral image classification","volume":"60","author":"Zou","year":"2022","journal-title":"IEEE Trans. Geosci. Remote Sens"}],"container-title":["Frontiers in Computer Science"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fcomp.2025.1658556\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,11,5]],"date-time":"2025-11-05T06:29:10Z","timestamp":1762324150000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fcomp.2025.1658556\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,11,5]]},"references-count":41,"alternative-id":["10.3389\/fcomp.2025.1658556"],"URL":"https:\/\/doi.org\/10.3389\/fcomp.2025.1658556","relation":{},"ISSN":["2624-9898"],"issn-type":[{"value":"2624-9898","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,11,5]]},"article-number":"1658556"}}