{"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/watergan-unsupervised-generative-network-to","title":"WaterGAN: Unsupervised Generative Network to Enable Real-time Color Correction of Monocular Underwater Images","arxiv_id":"1702.07392","date":"2017-02-23","proceeding":null,"authors":["Jie Li","Katherine A. Skinner","Ryan M. Eustice","Matthew Johnson-Roberson"],"abstract":"This paper reports on WaterGAN, a generative adversarial network (GAN) for\ngenerating realistic underwater images from in-air image and depth pairings in\nan unsupervised pipeline used for color correction of monocular underwater\nimages. Cameras onboard autonomous and remotely operated vehicles can capture\nhigh resolution images to map the seafloor, however, underwater image formation\nis subject to the complex process of light propagation through the water\ncolumn. The raw images retrieved are characteristically different than images\ntaken in air due to effects such as absorption and scattering, which cause\nattenuation of light at different rates for different wavelengths. While this\nphysical process is well described theoretically, the model depends on many\nparameters intrinsic to the water column as well as the objects in the scene.\nThese factors make recovery of these parameters difficult without simplifying\nassumptions or field calibration, hence, restoration of underwater images is a\nnon-trivial problem. Deep learning has demonstrated great success in modeling\ncomplex nonlinear systems but requires a large amount of training data, which\nis difficult to compile in deep sea environments. Using WaterGAN, we generate a\nlarge training dataset of paired imagery, both raw underwater and true color\nin-air, as well as depth data. This data serves as input to a novel end-to-end\nnetwork for color correction of monocular underwater images. Due to the\ndepth-dependent water column effects inherent to underwater environments, we\nshow that our end-to-end network implicitly learns a coarse depth estimate of\nthe underwater scene from monocular underwater images. Our proposed pipeline is\nvalidated with testing on real data collected from both a pure water tank and\nfrom underwater surveys in field testing. Source code is made publicly\navailable with sample datasets and pretrained models.","url_abs":"http://arxiv.org/abs/1702.07392v3","url_pdf":"http://arxiv.org/pdf/1702.07392v3.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":"watergan-unsupervised-generative-network-to","repo_url":"https://github.com/kskin/WaterGAN","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":null,"task_name":"Generative Adversarial Network"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1702.07392","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}