{"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/murel-multimodal-relational-reasoning-for","title":"MUREL: Multimodal Relational Reasoning for Visual Question Answering","arxiv_id":"1902.09487","date":"2019-02-25","proceeding":"CVPR 2019 6","authors":["Remi Cadene","Hedi Ben-Younes","Matthieu Cord","Nicolas Thome"],"abstract":"Multimodal attentional networks are currently state-of-the-art models for\nVisual Question Answering (VQA) tasks involving real images. Although attention\nallows to focus on the visual content relevant to the question, this simple\nmechanism is arguably insufficient to model complex reasoning features required\nfor VQA or other high-level tasks.\n  In this paper, we propose MuRel, a multimodal relational network which is\nlearned end-to-end to reason over real images. Our first contribution is the\nintroduction of the MuRel cell, an atomic reasoning primitive representing\ninteractions between question and image regions by a rich vectorial\nrepresentation, and modeling region relations with pairwise combinations.\nSecondly, we incorporate the cell into a full MuRel network, which\nprogressively refines visual and question interactions, and can be leveraged to\ndefine visualization schemes finer than mere attention maps.\n  We validate the relevance of our approach with various ablation studies, and\nshow its superiority to attention-based methods on three datasets: VQA 2.0,\nVQA-CP v2 and TDIUC. Our final MuRel network is competitive to or outperforms\nstate-of-the-art results in this challenging context.\n  Our code is available: https://github.com/Cadene/murel.bootstrap.pytorch","url_abs":"http://arxiv.org/abs/1902.09487v1","url_pdf":"http://arxiv.org/pdf/1902.09487v1.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":"murel-multimodal-relational-reasoning-for","repo_url":"https://github.com/Cadene/murel.bootstrap.pytorch","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"BSD-3-Clause"}}],"tasks":[{"task_slug":"relational-reasoning","task_name":"Relational Reasoning"},{"task_slug":"visual-question-answering-1","task_name":"Visual Question Answering"},{"task_slug":"visual-question-answering","task_name":"Visual Question Answering (VQA)"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/visual-question-answering-on-tdiuc","task":"Visual Question Answering (VQA)","dataset":"TDIUC","model":"Accuracy","rank_in_archive_order":1,"of":2,"metrics":{"Accuracy":"88.2"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-on-vqa-v2-test-dev","task":"Visual Question Answering (VQA)","dataset":"VQA v2 test-dev","model":"MuRel","rank_in_archive_order":37,"of":56,"metrics":{"Accuracy":"68.03"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-on-vqa-v2-test-std","task":"Visual Question Answering (VQA)","dataset":"VQA v2 test-std","model":"MuRel","rank_in_archive_order":31,"of":38,"metrics":{"overall":"68.4"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-on-vqa-cp","task":"Visual Question Answering (VQA)","dataset":"VQA-CP","model":"MuRel","rank_in_archive_order":9,"of":10,"metrics":{"Score":"39.54"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1902.09487","atlas_url":"https://app.syntology.ai/?focus=1902.09487","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}