{"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/the-effects-of-super-resolution-on-object","title":"The Effects of Super-Resolution on Object Detection Performance in Satellite Imagery","arxiv_id":"1812.04098","date":"2018-12-10","proceeding":null,"authors":["Jacob Shermeyer","Adam Van Etten"],"abstract":"We explore the application of super-resolution techniques to satellite\nimagery, and the effects of these techniques on object detection algorithm\nperformance. Specifically, we enhance satellite imagery beyond its native\nresolution, and test if we can identify various types of vehicles, planes, and\nboats with greater accuracy than native resolution. Using the Very Deep\nSuper-Resolution (VDSR) framework and a custom Random Forest Super-Resolution\n(RFSR) framework we generate enhancement levels of 2x, 4x, and 8x over five\ndistinct resolutions ranging from 30 cm to 4.8 meters. Using both native and\nsuper-resolved data, we then train several custom detection models using the\nSIMRDWN object detection framework. SIMRDWN combines a number of popular object\ndetection algorithms (e.g. SSD, YOLO) into a unified framework that is designed\nto rapidly detect objects in large satellite images. This approach allows us to\nquantify the effects of super-resolution techniques on object detection\nperformance across multiple classes and resolutions. We also quantify the\nperformance of object detection as a function of native resolution and object\npixel size. For our test set we note that performance degrades from mean\naverage precision (mAP) = 0.53 at 30 cm resolution, down to mAP = 0.11 at 4.8 m\nresolution. Super-resolving native 30 cm imagery to 15 cm yields the greatest\nbenefit; a 13-36% improvement in mAP. Super-resolution is less beneficial at\ncoarser resolutions, though still provides a small improvement in performance.","url_abs":"http://arxiv.org/abs/1812.04098v3","url_pdf":"http://arxiv.org/pdf/1812.04098v3.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":"the-effects-of-super-resolution-on-object","repo_url":"https://github.com/jshermeyer/RFSR","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"the-effects-of-super-resolution-on-object","repo_url":"https://github.com/jshermeyer/VDSR4Geo","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"object","task_name":"Object"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"super-resolution","task_name":"Super-Resolution"},{"task_slug":"object-detection-1","task_name":"object-detection"},{"task_slug":"satellite-image-super-resolution","task_name":"satellite image super-resolution"}],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"non-maximum-suppression","method_name":"Non Maximum Suppression"},{"method_slug":"ssd","method_name":"SSD"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1812.04098","atlas_url":"https://app.syntology.ai/?focus=1812.04098","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}