{"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/sketchmate-deep-hashing-for-million-scale","title":"SketchMate: Deep Hashing for Million-Scale Human Sketch Retrieval","arxiv_id":"1804.01401","date":"2018-04-04","proceeding":"CVPR 2018 6","authors":["Peng Xu","Yongye Huang","Tongtong Yuan","Kaiyue Pang","Yi-Zhe Song","Tao Xiang","Timothy M. Hospedales","Zhanyu Ma","Jun Guo"],"abstract":"We propose a deep hashing framework for sketch retrieval that, for the first\ntime, works on a multi-million scale human sketch dataset. Leveraging on this\nlarge dataset, we explore a few sketch-specific traits that were otherwise\nunder-studied in prior literature. Instead of following the conventional sketch\nrecognition task, we introduce the novel problem of sketch hashing retrieval\nwhich is not only more challenging, but also offers a better testbed for\nlarge-scale sketch analysis, since: (i) more fine-grained sketch feature\nlearning is required to accommodate the large variations in style and\nabstraction, and (ii) a compact binary code needs to be learned at the same\ntime to enable efficient retrieval. Key to our network design is the embedding\nof unique characteristics of human sketch, where (i) a two-branch CNN-RNN\narchitecture is adapted to explore the temporal ordering of strokes, and (ii) a\nnovel hashing loss is specifically designed to accommodate both the temporal\nand abstract traits of sketches. By working with a 3.8M sketch dataset, we show\nthat state-of-the-art hashing models specifically engineered for static images\nfail to perform well on temporal sketch data. Our network on the other hand not\nonly offers the best retrieval performance on various code sizes, but also\nyields the best generalization performance under a zero-shot setting and when\nre-purposed for sketch recognition. Such superior performances effectively\ndemonstrate the benefit of our sketch-specific design.","url_abs":"http://arxiv.org/abs/1804.01401v1","url_pdf":"http://arxiv.org/pdf/1804.01401v1.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":"sketchmate-deep-hashing-for-million-scale","repo_url":"https://github.com/tosmaster/imagevision","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"deep-hashing","task_name":"Deep Hashing"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"sketch-recognition","task_name":"Sketch Recognition"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1804.01401","atlas_url":"https://app.syntology.ai/?focus=1804.01401","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}