{"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/a-zero-shot-framework-for-sketch-based-image","title":"A Zero-Shot Framework for Sketch-based Image Retrieval","arxiv_id":"1807.11724","date":"2018-07-31","proceeding":null,"authors":["Sasi Kiran Yelamarthi","Shiva Krishna Reddy","Ashish Mishra","Anurag Mittal"],"abstract":"Sketch-based image retrieval (SBIR) is the task of retrieving images from a\nnatural image database that correspond to a given hand-drawn sketch. Ideally,\nan SBIR model should learn to associate components in the sketch (say, feet,\ntail, etc.) with the corresponding components in the image having similar shape\ncharacteristics. However, current evaluation methods simply focus only on\ncoarse-grained evaluation where the focus is on retrieving images which belong\nto the same class as the sketch but not necessarily having the same shape\ncharacteristics as in the sketch. As a result, existing methods simply learn to\nassociate sketches with classes seen during training and hence fail to\ngeneralize to unseen classes. In this paper, we propose a new benchmark for\nzero-shot SBIR where the model is evaluated in novel classes that are not seen\nduring training. We show through extensive experiments that existing models for\nSBIR that are trained in a discriminative setting learn only class specific\nmappings and fail to generalize to the proposed zero-shot setting. To\ncircumvent this, we propose a generative approach for the SBIR task by\nproposing deep conditional generative models that take the sketch as an input\nand fill the missing information stochastically. Experiments on this new\nbenchmark created from the \"Sketchy\" dataset, which is a large-scale database\nof sketch-photo pairs demonstrate that the performance of these generative\nmodels is significantly better than several state-of-the-art approaches in the\nproposed zero-shot framework of the coarse-grained SBIR task.","url_abs":"http://arxiv.org/abs/1807.11724v1","url_pdf":"http://arxiv.org/pdf/1807.11724v1.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":"a-zero-shot-framework-for-sketch-based-image","repo_url":"https://github.com/ShivaKrishnaM/ZS-SBIR","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"image-retrieval","task_name":"Image Retrieval"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"sketch-based-image-retrieval","task_name":"Sketch-Based Image Retrieval"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1807.11724","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}