{"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/towards-multi-class-object-detection-in","title":"Towards Multi-class Object Detection in Unconstrained Remote Sensing Imagery","arxiv_id":"1807.02700","date":"2018-07-07","proceeding":null,"authors":["Seyed Majid Azimi","Eleonora Vig","Reza Bahmanyar","Marco Körner","Peter Reinartz"],"abstract":"Automatic multi-class object detection in remote sensing images in\nunconstrained scenarios is of high interest for several applications including\ntraffic monitoring and disaster management. The huge variation in object scale,\norientation, category, and complex backgrounds, as well as the different camera\nsensors pose great challenges for current algorithms. In this work, we propose\na new method consisting of a novel joint image cascade and feature pyramid\nnetwork with multi-size convolution kernels to extract multi-scale strong and\nweak semantic features. These features are fed into rotation-based region\nproposal and region of interest networks to produce object detections. Finally,\nrotational non-maximum suppression is applied to remove redundant detections.\nDuring training, we minimize joint horizontal and oriented bounding box loss\nfunctions, as well as a novel loss that enforces oriented boxes to be\nrectangular. Our method achieves 68.16% mAP on horizontal and 72.45% mAP on\noriented bounding box detection tasks on the challenging DOTA dataset,\noutperforming all published methods by a large margin (+6% and +12% absolute\nimprovement, respectively). Furthermore, it generalizes to two other datasets,\nNWPU VHR-10 and UCAS-AOD, and achieves competitive results with the baselines\neven when trained on DOTA. Our method can be deployed in multi-class object\ndetection applications, regardless of the image and object scales and\norientations, making it a great choice for unconstrained aerial and satellite\nimagery.","url_abs":"http://arxiv.org/abs/1807.02700v3","url_pdf":"http://arxiv.org/pdf/1807.02700v3.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":[],"tasks":[{"task_slug":"management","task_name":"Management"},{"task_slug":"object","task_name":"Object"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"object-detection-in-aerial-images","task_name":"Object Detection In Aerial Images"},{"task_slug":"region-proposal","task_name":"Region Proposal"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/object-detection-in-aerial-images-on-dota-1","task":"Object Detection In Aerial Images","dataset":"DOTA","model":"ICN","rank_in_archive_order":55,"of":58,"metrics":{"mAP":"68.16%"},"uses_additional_data":true}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1807.02700","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}