{"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/learning-token-based-representation-for-image","title":"Learning Token-based Representation for Image Retrieval","arxiv_id":"2112.06159","date":"2021-12-12","proceeding":null,"authors":["Hui Wu","Min Wang","Wengang Zhou","Yang Hu","Houqiang Li"],"abstract":"In image retrieval, deep local features learned in a data-driven manner have been demonstrated effective to improve retrieval performance. To realize efficient retrieval on large image database, some approaches quantize deep local features with a large codebook and match images with aggregated match kernel. However, the complexity of these approaches is non-trivial with large memory footprint, which limits their capability to jointly perform feature learning and aggregation. To generate compact global representations while maintaining regional matching capability, we propose a unified framework to jointly learn local feature representation and aggregation. In our framework, we first extract deep local features using CNNs. Then, we design a tokenizer module to aggregate them into a few visual tokens, each corresponding to a specific visual pattern. This helps to remove background noise, and capture more discriminative regions in the image. Next, a refinement block is introduced to enhance the visual tokens with self-attention and cross-attention. Finally, different visual tokens are concatenated to generate a compact global representation. The whole framework is trained end-to-end with image-level labels. Extensive experiments are conducted to evaluate our approach, which outperforms the state-of-the-art methods on the Revisited Oxford and Paris datasets.","url_abs":"https://arxiv.org/abs/2112.06159v1","url_pdf":"https://arxiv.org/pdf/2112.06159v1.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":"learning-token-based-representation-for-image","repo_url":"https://github.com/mcc-wh/token","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"image-retrieval","task_name":"Image Retrieval"},{"task_slug":"retrieval","task_name":"Retrieval"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-retrieval-on-roxford-hard","task":"Image Retrieval","dataset":"ROxford (Hard)","model":"Token","rank_in_archive_order":4,"of":23,"metrics":{"mAP":"66.57"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-roxford-medium","task":"Image Retrieval","dataset":"ROxford (Medium)","model":"Token","rank_in_archive_order":3,"of":23,"metrics":{"mAP":"82.28"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-hard","task":"Image Retrieval","dataset":"RParis (Hard)","model":"Token","rank_in_archive_order":4,"of":23,"metrics":{"mAP":"78.56"},"uses_additional_data":false},{"leaderboard":"/sota/image-retrieval-on-rparis-medium","task":"Image Retrieval","dataset":"RParis (Medium)","model":"Token","rank_in_archive_order":3,"of":23,"metrics":{"mAP":"89.34"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=2112.06159","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}