{"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/large-scale-long-tailed-recognition-in-an","title":"Large-Scale Long-Tailed Recognition in an Open World","arxiv_id":"1904.05160","date":"2019-04-10","proceeding":"CVPR 2019 6","authors":["Ziwei Liu","Zhongqi Miao","Xiaohang Zhan","Jiayun Wang","Boqing Gong","Stella X. Yu"],"abstract":"Real world data often have a long-tailed and open-ended distribution. A\npractical recognition system must classify among majority and minority classes,\ngeneralize from a few known instances, and acknowledge novelty upon a never\nseen instance. We define Open Long-Tailed Recognition (OLTR) as learning from\nsuch naturally distributed data and optimizing the classification accuracy over\na balanced test set which include head, tail, and open classes. OLTR must\nhandle imbalanced classification, few-shot learning, and open-set recognition\nin one integrated algorithm, whereas existing classification approaches focus\nonly on one aspect and deliver poorly over the entire class spectrum. The key\nchallenges are how to share visual knowledge between head and tail classes and\nhow to reduce confusion between tail and open classes. We develop an integrated\nOLTR algorithm that maps an image to a feature space such that visual concepts\ncan easily relate to each other based on a learned metric that respects the\nclosed-world classification while acknowledging the novelty of the open world.\nOur so-called dynamic meta-embedding combines a direct image feature and an\nassociated memory feature, with the feature norm indicating the familiarity to\nknown classes. On three large-scale OLTR datasets we curate from object-centric\nImageNet, scene-centric Places, and face-centric MS1M data, our method\nconsistently outperforms the state-of-the-art. Our code, datasets, and models\nenable future OLTR research and are publicly available at\nhttps://liuziwei7.github.io/projects/LongTail.html.","url_abs":"http://arxiv.org/abs/1904.05160v2","url_pdf":"http://arxiv.org/pdf/1904.05160v2.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":"large-scale-long-tailed-recognition-in-an","repo_url":"https://github.com/silicx/dlsa","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"large-scale-long-tailed-recognition-in-an","repo_url":"https://github.com/zhmiao/OpenLongTailRecognition-OLTR","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"BSD-3-Clause"}}],"tasks":[{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"few-shot-learning","task_name":"Few-Shot Learning"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"long-tail-learning","task_name":"Long-tail Learning"},{"task_slug":"long-tail-learning-with-class-descriptors","task_name":"Long-tail learning with class descriptors"},{"task_slug":"open-set-learning","task_name":"Open Set Learning"},{"task_slug":"imbalanced-classification","task_name":"imbalanced classification"}],"methods":[],"datasets_introduced":[{"slug":"imagenet-lt","name":"ImageNet-LT","full_name":"ImageNet Long-Tailed"},{"slug":"places-lt","name":"Places-LT","full_name":""}],"methods_introduced":[],"results":[{"leaderboard":"/sota/long-tail-learning-on-coco-mlt","task":"Long-tail Learning","dataset":"COCO-MLT","model":"OLTR(ResNet-50)","rank_in_archive_order":11,"of":13,"metrics":{"Average mAP":"45.83"},"uses_additional_data":false},{"leaderboard":"/sota/long-tail-learning-on-imagenet-lt","task":"Long-tail Learning","dataset":"ImageNet-LT","model":"OLTR","rank_in_archive_order":67,"of":69,"metrics":{"Top-1 Accuracy":"35.6"},"uses_additional_data":false},{"leaderboard":"/sota/long-tail-learning-on-places-lt","task":"Long-tail Learning","dataset":"Places-LT","model":"OLTR","rank_in_archive_order":28,"of":29,"metrics":{"Top-1 Accuracy":"34.1"},"uses_additional_data":false},{"leaderboard":"/sota/long-tail-learning-on-voc-mlt","task":"Long-tail Learning","dataset":"VOC-MLT","model":"OLTR(ResNet-50)","rank_in_archive_order":11,"of":13,"metrics":{"Average mAP":"71.02"},"uses_additional_data":false},{"leaderboard":"/sota/long-tail-learning-with-class-descriptors-on-3","task":"Long-tail learning with class descriptors","dataset":"ImageNet-LT-d","model":"OLTR","rank_in_archive_order":5,"of":5,"metrics":{"Per-Class Accuracy":"37.7"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/1904.05160","atlas_url":"https://app.syntology.ai/?focus=1904.05160","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}