{"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/efficient-interactive-annotation-of","title":"Efficient Interactive Annotation of Segmentation Datasets with Polygon-RNN++","arxiv_id":"1803.09693","date":"2018-03-26","proceeding":"CVPR 2018 6","authors":["David Acuna","Huan Ling","Amlan Kar","Sanja Fidler"],"abstract":"Manually labeling datasets with object masks is extremely time consuming. In\nthis work, we follow the idea of Polygon-RNN to produce polygonal annotations\nof objects interactively using humans-in-the-loop. We introduce several\nimportant improvements to the model: 1) we design a new CNN encoder\narchitecture, 2) show how to effectively train the model with Reinforcement\nLearning, and 3) significantly increase the output resolution using a Graph\nNeural Network, allowing the model to accurately annotate high-resolution\nobjects in images. Extensive evaluation on the Cityscapes dataset shows that\nour model, which we refer to as Polygon-RNN++, significantly outperforms the\noriginal model in both automatic (10% absolute and 16% relative improvement in\nmean IoU) and interactive modes (requiring 50% fewer clicks by annotators). We\nfurther analyze the cross-domain scenario in which our model is trained on one\ndataset, and used out of the box on datasets from varying domains. The results\nshow that Polygon-RNN++ exhibits powerful generalization capabilities,\nachieving significant improvements over existing pixel-wise methods. Using\nsimple online fine-tuning we further achieve a high reduction in annotation\ntime for new datasets, moving a step closer towards an interactive annotation\ntool to be used in practice.","url_abs":"http://arxiv.org/abs/1803.09693v1","url_pdf":"http://arxiv.org/pdf/1803.09693v1.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":"efficient-interactive-annotation-of","repo_url":"https://github.com/AidanRocke/vertex_prediction","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}},{"paper_slug":"efficient-interactive-annotation-of","repo_url":"https://github.com/fidler-lab/polyrnn-pp","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"GPL-3.0"}},{"paper_slug":"efficient-interactive-annotation-of","repo_url":"https://github.com/fidler-lab/polyrnn-pp-pytorch","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"graph-neural-network","task_name":"Graph Neural Network"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1803.09693","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}