{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/dynamic-multi-scale-segmentation-of-remote","title":"Dynamic Multi-Context Segmentation of Remote Sensing Images based on Convolutional Networks","arxiv_id":"1804.04020","date":"2018-04-11","proceeding":null,"authors":["Keiller Nogueira","Mauro Dalla Mura","Jocelyn Chanussot","William R. Schwartz","Jefersson A. dos Santos"],"abstract":"Semantic segmentation requires methods capable of learning high-level\nfeatures while dealing with large volume of data. Towards such goal,\nConvolutional Networks can learn specific and adaptable features based on the\ndata. However, these networks are not capable of processing a whole remote\nsensing image, given its huge size. To overcome such limitation, the image is\nprocessed using fixed size patches. The definition of the input patch size is\nusually performed empirically (evaluating several sizes) or imposed (by network\nconstraint). Both strategies suffer from drawbacks and could not lead to the\nbest patch size. To alleviate this problem, several works exploited\nmulti-context information by combining networks or layers. This process\nincreases the number of parameters resulting in a more difficult model to\ntrain. In this work, we propose a novel technique to perform semantic\nsegmentation of remote sensing images that exploits a multi-context paradigm\nwithout increasing the number of parameters while defining, in training time,\nthe best patch size. The main idea is to train a dilated network with distinct\npatch sizes, allowing it to capture multi-context characteristics from\nheterogeneous contexts. While processing these varying patches, the network\nprovides a score for each patch size, helping in the definition of the best\nsize for the current scenario. A systematic evaluation of the proposed\nalgorithm is conducted using four high-resolution remote sensing datasets with\nvery distinct properties. Our results show that the proposed algorithm provides\nimprovements in pixelwise classification accuracy when compared to\nstate-of-the-art methods.","url_abs":"http://arxiv.org/abs/1804.04020v3","url_pdf":"http://arxiv.org/pdf/1804.04020v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"dynamic-multi-scale-segmentation-of-remote","repo_url":"https://github.com/keillernogueira/dynamic-rs-segmentation","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}