{"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-real-time-automatic-portrait-matting","title":"Towards Real-Time Automatic Portrait Matting on Mobile Devices","arxiv_id":"1904.03816","date":"2019-04-08","proceeding":null,"authors":["Seokjun Seo","Seungwoo Choi","Martin Kersner","Beomjun Shin","Hyungsuk Yoon","Hyeongmin Byun","Sungjoo Ha"],"abstract":"We tackle the problem of automatic portrait matting on mobile devices. The\nproposed model is aimed at attaining real-time inference on mobile devices with\nminimal degradation of model performance. Our model MMNet, based on\nmulti-branch dilated convolution with linear bottleneck blocks, outperforms the\nstate-of-the-art model and is orders of magnitude faster. The model can be\naccelerated four times to attain 30 FPS on Xiaomi Mi 5 device with moderate\nincrease in the gradient error. Under the same conditions, our model has an\norder of magnitude less number of parameters and is faster than Mobile\nDeepLabv3 while maintaining comparable performance. The accompanied\nimplementation can be found at \\url{https://github.com/hyperconnect/MMNet}.","url_abs":"http://arxiv.org/abs/1904.03816v1","url_pdf":"http://arxiv.org/pdf/1904.03816v1.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":"towards-real-time-automatic-portrait-matting","repo_url":"https://github.com/hyperconnect/MMNet","is_official":0,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":null},{"paper_slug":"towards-real-time-automatic-portrait-matting","repo_url":"https://github.com/minus31/Portrait-segmentation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"image-matting","task_name":"Image Matting"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dilated-convolution","method_name":"Dilated Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1904.03816","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}