{"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/florence-a-new-foundation-model-for-computer","title":"Florence: A New Foundation Model for Computer Vision","arxiv_id":"2111.11432","date":"2021-11-22","proceeding":null,"authors":["Lu Yuan","Dongdong Chen","Yi-Ling Chen","Noel Codella","Xiyang Dai","Jianfeng Gao","Houdong Hu","Xuedong Huang","Boxin Li","Chunyuan Li","Ce Liu","Mengchen Liu","Zicheng Liu","Yumao Lu","Yu Shi","Lijuan Wang","JianFeng Wang","Bin Xiao","Zhen Xiao","Jianwei Yang","Michael Zeng","Luowei Zhou","Pengchuan Zhang"],"abstract":"Automated visual understanding of our diverse and open world demands computer vision models to generalize well with minimal customization for specific tasks, similar to human vision. Computer vision foundation models, which are trained on diverse, large-scale dataset and can be adapted to a wide range of downstream tasks, are critical for this mission to solve real-world computer vision applications. While existing vision foundation models such as CLIP, ALIGN, and Wu Dao 2.0 focus mainly on mapping images and textual representations to a cross-modal shared representation, we introduce a new computer vision foundation model, Florence, to expand the representations from coarse (scene) to fine (object), from static (images) to dynamic (videos), and from RGB to multiple modalities (caption, depth). By incorporating universal visual-language representations from Web-scale image-text data, our Florence model can be easily adapted for various computer vision tasks, such as classification, retrieval, object detection, VQA, image caption, video retrieval and action recognition. Moreover, Florence demonstrates outstanding performance in many types of transfer learning: fully sampled fine-tuning, linear probing, few-shot transfer and zero-shot transfer for novel images and objects. All of these properties are critical for our vision foundation model to serve general purpose vision tasks. Florence achieves new state-of-the-art results in majority of 44 representative benchmarks, e.g., ImageNet-1K zero-shot classification with top-1 accuracy of 83.74 and the top-5 accuracy of 97.18, 62.4 mAP on COCO fine tuning, 80.36 on VQA, and 87.8 on Kinetics-600.","url_abs":"https://arxiv.org/abs/2111.11432v1","url_pdf":"https://arxiv.org/pdf/2111.11432v1.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":"florence-a-new-foundation-model-for-computer","repo_url":"https://github.com/microsoft/unicl","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"florence-a-new-foundation-model-for-computer","repo_url":"https://github.com/MindCode-4/code-3/tree/main/florence2","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null}],"tasks":[{"task_slug":"action-classification","task_name":"Action Classification"},{"task_slug":"action-recognition-in-videos","task_name":"Action Recognition"},{"task_slug":"action-recognition-in-videos-2","task_name":"Action Recognition In Videos"},{"task_slug":"cross-modal-retrieval","task_name":"Cross-Modal Retrieval"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"retrieval","task_name":"Retrieval"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"},{"task_slug":"video-retrieval","task_name":"Video Retrieval"},{"task_slug":"visual-question-answering-1","task_name":"Visual Question Answering"},{"task_slug":"visual-question-answering","task_name":"Visual Question Answering (VQA)"},{"task_slug":"zero-shot-cross-modal-retrieval","task_name":"Zero-Shot Cross-Modal Retrieval"},{"task_slug":"zero-shot-learning","task_name":"Zero-Shot Learning"},{"task_slug":"zero-shot-transfer-image-classification","task_name":"Zero-Shot Transfer Image Classification"},{"task_slug":"zero-shot-transfer-image-classification-cn","task_name":"Zero-Shot Transfer Image Classification (CN)"},{"task_slug":"zero-shot-video-retrieval","task_name":"Zero-Shot Video Retrieval"},{"task_slug":"model","task_name":"model"},{"task_slug":"object-detection-1","task_name":"object-detection"},{"task_slug":null,"task_name":"zero-shot-classification"}],"methods":[{"method_slug":"align","method_name":"ALIGN"},{"method_slug":"clip","method_name":"CLIP"},{"method_slug":"florence","method_name":"Florence"}],"datasets_introduced":[],"methods_introduced":[{"slug":"florence","name":"Florence","full_name":"Florence"}],"results":[{"leaderboard":"/sota/action-classification-on-kinetics-600","task":"Action Classification","dataset":"Kinetics-600","model":"Florence (curated FLD-900M pretrain)","rank_in_archive_order":25,"of":65,"metrics":{"Top-1 Accuracy":"87.8","Top-5 Accuracy":"97.9"},"uses_additional_data":true},{"leaderboard":"/sota/action-recognition-in-videos-on-kinetics-400-1","task":"Action Recognition In Videos","dataset":"Kinetics-400","model":"Florence","rank_in_archive_order":1,"of":3,"metrics":{"Top-1 Accuracy":"86.5","Top-5 Accuracy":"97.3"},"uses_additional_data":false},{"leaderboard":"/sota/action-recognition-in-videos-on-kinetics-600","task":"Action Recognition In Videos","dataset":"Kinetics-600","model":"Florence","rank_in_archive_order":1,"of":1,"metrics":{"Top-1 Accuracy":"87.8","Top-5 Accuracy":"97.8"},"uses_additional_data":false},{"leaderboard":"/sota/cross-modal-retrieval-on-coco-2014","task":"Cross-Modal Retrieval","dataset":"COCO 2014","model":"Florence","rank_in_archive_order":10,"of":36,"metrics":{"Image-to-text R@1":"81.8","Image-to-text R@5":"95.2","Text-to-image R@1":"63.2","Text-to-image R@5":"85.7"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"Florence-CoSwin-H","rank_in_archive_order":9,"of":1060,"metrics":{"Number of params":"893M","Top 1 Accuracy":"90.05%","Top 5 Accuracy":"99.02"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco-minival","task":"Object Detection","dataset":"COCO minival","model":"Florence-CoSwin-H","rank_in_archive_order":16,"of":220,"metrics":{"box AP":"62"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-coco","task":"Object Detection","dataset":"COCO test-dev","model":"Florence-CoSwin-H","rank_in_archive_order":20,"of":225,"metrics":{"box mAP":"62.4"},"uses_additional_data":false},{"leaderboard":"/sota/video-retrieval-on-msr-vtt-1ka","task":"Video Retrieval","dataset":"MSR-VTT-1kA","model":"Florence","rank_in_archive_order":44,"of":63,"metrics":{"text-to-video R@1":"37.6","text-to-video R@10":"72.6","text-to-video R@5":"63.8"},"uses_additional_data":true},{"leaderboard":"/sota/visual-question-answering-on-vqa-v2-test-dev-1","task":"Visual Question Answering","dataset":"VQA v2 test-dev","model":"Florence","rank_in_archive_order":7,"of":11,"metrics":{"Accuracy":"80.16"},"uses_additional_data":false},{"leaderboard":"/sota/visual-question-answering-on-vqa-v2-test-std-1","task":"Visual Question Answering","dataset":"VQA v2 test-std","model":"Florence","rank_in_archive_order":3,"of":3,"metrics":{"overall":"80.36"},"uses_additional_data":false},{"leaderboard":"/sota/zero-shot-cross-modal-retrieval-on-coco-2014","task":"Zero-Shot Cross-Modal Retrieval","dataset":"COCO 2014","model":"Florence","rank_in_archive_order":11,"of":18,"metrics":{"Image-to-text R@1":"64.7","Image-to-text R@5":"85.9","Text-to-image R@1":"47.2","Text-to-image R@5":"71.4"},"uses_additional_data":false},{"leaderboard":"/sota/zero-shot-cross-modal-retrieval-on-flickr30k","task":"Zero-Shot Cross-Modal Retrieval","dataset":"Flickr30k","model":"Florence","rank_in_archive_order":9,"of":22,"metrics":{"Image-to-text R@1":"90.9","Image-to-text R@10":"-","Image-to-text R@5":"99.1","Text-to-image R@1":"76.7","Text-to-image R@10":"-","Text-to-image R@5":"93.6"},"uses_additional_data":false},{"leaderboard":"/sota/zero-shot-video-retrieval-on-msr-vtt","task":"Zero-Shot Video Retrieval","dataset":"MSR-VTT","model":"Florence","rank_in_archive_order":15,"of":41,"metrics":{"text-to-video R@1":"37.6","text-to-video R@10":"72.6","text-to-video R@5":"63.8"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2111.11432","atlas_url":"https://app.syntology.ai/?focus=2111.11432","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}