{"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/structured-label-inference-for-visual","title":"Structured Label Inference for Visual Understanding","arxiv_id":"1802.06459","date":"2018-02-18","proceeding":null,"authors":["Nelson Nauata","Hexiang Hu","Guang-Tong Zhou","Zhiwei Deng","Zicheng Liao","Greg Mori"],"abstract":"Visual data such as images and videos contain a rich source of structured\nsemantic labels as well as a wide range of interacting components. Visual\ncontent could be assigned with fine-grained labels describing major components,\ncoarse-grained labels depicting high level abstractions, or a set of labels\nrevealing attributes. Such categorization over different, interacting layers of\nlabels evinces the potential for a graph-based encoding of label information.\nIn this paper, we exploit this rich structure for performing graph-based\ninference in label space for a number of tasks: multi-label image and video\nclassification and action detection in untrimmed videos. We consider the use of\nthe Bidirectional Inference Neural Network (BINN) and Structured Inference\nNeural Network (SINN) for performing graph-based inference in label space and\npropose a Long Short-Term Memory (LSTM) based extension for exploiting activity\nprogression on untrimmed videos. The methods were evaluated on (i) the Animal\nwith Attributes (AwA), Scene Understanding (SUN) and NUS-WIDE datasets for\nmulti-label image classification, (ii) the first two releases of the YouTube-8M\nlarge scale dataset for multi-label video classification, and (iii) the\nTHUMOS'14 and MultiTHUMOS video datasets for action detection. Our results\ndemonstrate the effectiveness of structured label inference in these\nchallenging tasks, achieving significant improvements against baselines.","url_abs":"http://arxiv.org/abs/1802.06459v1","url_pdf":"http://arxiv.org/pdf/1802.06459v1.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":"structured-label-inference-for-visual","repo_url":"https://github.com/daveboat/structured_label_inference","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"action-detection","task_name":"Action Detection"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"multi-label-image-classification","task_name":"Multi-Label Image Classification"},{"task_slug":"scene-understanding","task_name":"Scene Understanding"},{"task_slug":"video-classification","task_name":"Video Classification"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1802.06459","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}