{"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/eva-exploring-the-limits-of-masked-visual","title":"EVA: Exploring the Limits of Masked Visual Representation Learning at Scale","arxiv_id":"2211.07636","date":"2022-11-14","proceeding":"CVPR 2023 1","authors":["Yuxin Fang","Wen Wang","Binhui Xie","Quan Sun","Ledell Wu","Xinggang Wang","Tiejun Huang","Xinlong Wang","Yue Cao"],"abstract":"We launch EVA, a vision-centric foundation model to explore the limits of visual representation at scale using only publicly accessible data. EVA is a vanilla ViT pre-trained to reconstruct the masked out image-text aligned vision features conditioned on visible image patches. Via this pretext task, we can efficiently scale up EVA to one billion parameters, and sets new records on a broad range of representative vision downstream tasks, such as image recognition, video action recognition, object detection, instance segmentation and semantic segmentation without heavy supervised training. Moreover, we observe quantitative changes in scaling EVA result in qualitative changes in transfer learning performance that are not present in other models. For instance, EVA takes a great leap in the challenging large vocabulary instance segmentation task: our model achieves almost the same state-of-the-art performance on LVISv1.0 dataset with over a thousand categories and COCO dataset with only eighty categories. Beyond a pure vision encoder, EVA can also serve as a vision-centric, multi-modal pivot to connect images and text. We find initializing the vision tower of a giant CLIP from EVA can greatly stabilize the training and outperform the training from scratch counterpart with much fewer samples and less compute, providing a new direction for scaling up and accelerating the costly training of multi-modal foundation models. To facilitate future research, we release all the code and models at https://github.com/baaivision/EVA.","url_abs":"https://arxiv.org/abs/2211.07636v2","url_pdf":"https://arxiv.org/pdf/2211.07636v2.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":"eva-exploring-the-limits-of-masked-visual","repo_url":"https://github.com/baaivision/eva","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"eva-exploring-the-limits-of-masked-visual","repo_url":"https://github.com/rwightman/pytorch-image-models","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"eva-exploring-the-limits-of-masked-visual","repo_url":"https://github.com/PaddlePaddle/PASSL","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"paddle","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"eva-exploring-the-limits-of-masked-visual","repo_url":"https://github.com/PaddlePaddle/PaddleMIX/tree/develop/paddlemix/examples","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"paddle","reach":null},{"paper_slug":"eva-exploring-the-limits-of-masked-visual","repo_url":"https://github.com/leondgarse/keras_cv_attention_models/tree/main/keras_cv_attention_models/beit","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null},{"paper_slug":"eva-exploring-the-limits-of-masked-visual","repo_url":"https://github.com/open-mmlab/mmselfsup","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"action-classification","task_name":"Action Classification"},{"task_slug":"action-recognition-in-videos","task_name":"Action Recognition"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"instance-segmentation","task_name":"Instance Segmentation"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"representation-learning","task_name":"Representation Learning"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"self-supervised-image-classification","task_name":"Self-Supervised Image Classification"},{"task_slug":null,"task_name":"Self-Supervised Image Classification (with CLIP)"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"action-recognition","task_name":"Temporal Action Localization"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"}],"methods":[{"method_slug":"clip","method_name":"CLIP"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/action-classification-on-kinetics-400","task":"Action Classification","dataset":"Kinetics-400","model":"EVA","rank_in_archive_order":16,"of":207,"metrics":{"Acc@1":"89.7"},"uses_additional_data":false},{"leaderboard":"/sota/action-classification-on-kinetics-600","task":"Action Classification","dataset":"Kinetics-600","model":"EVA","rank_in_archive_order":13,"of":65,"metrics":{"Top-1 Accuracy":"89.8%"},"uses_additional_data":true},{"leaderboard":"/sota/action-classification-on-kinetics-700","task":"Action Classification","dataset":"Kinetics-700","model":"EVA","rank_in_archive_order":7,"of":36,"metrics":{"Top-1 Accuracy":"82.9%"},"uses_additional_data":true},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"EVA","rank_in_archive_order":13,"of":1060,"metrics":{"Number of params":"1000M","Top 1 Accuracy":"89.7%"},"uses_additional_data":false},{"leaderboard":"/sota/instance-segmentation-on-coco-minival","task":"Instance Segmentation","dataset":"COCO minival","model":"EVA","rank_in_archive_order":4,"of":93,"metrics":{"AP50":"79.4","AP75":"60.9","APL":"72.0","APM":"58.4","APS":"37.6","mask AP":"55.0"},"uses_additional_data":true},{"leaderboard":"/sota/instance-segmentation-on-coco","task":"Instance Segmentation","dataset":"COCO test-dev","model":"EVA","rank_in_archive_order":3,"of":112,"metrics":{"AP50":"80.0","APL":"72.4","APM":"58.0","APS":"36.3","mask AP":"55.5"},"uses_additional_data":true},{"leaderboard":"/sota/instance-segmentation-on-lvis-v1-0-val","task":"Instance Segmentation","dataset":"LVIS v1.0 val","model":"EVA","rank_in_archive_order":2,"of":25,"metrics":{"mask AP":"55.0"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-coco-minival","task":"Object Detection","dataset":"COCO minival","model":"EVA","rank_in_archive_order":7,"of":220,"metrics":{"AP50":"82.1","AP75":"70.8","APL":"78.5","APM":"68.4","APS":"49.4","box AP":"64.5"},"uses_additional_data":true},{"leaderboard":"/sota/object-detection-on-coco","task":"Object Detection","dataset":"COCO test-dev","model":"EVA","rank_in_archive_order":7,"of":225,"metrics":{"AP50":"81.9","AP75":"71.7","APL":"77.9","APM":"67.7","APS":"48.5","box mAP":"64.7"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-coco-o","task":"Object Detection","dataset":"COCO-O","model":"EVA","rank_in_archive_order":1,"of":45,"metrics":{"Average mAP":"57.8","Effective Robustness":"28.86"},"uses_additional_data":false},{"leaderboard":"/sota/object-detection-on-lvis-v1-0-val","task":"Object Detection","dataset":"LVIS v1.0 val","model":"EVA","rank_in_archive_order":4,"of":15,"metrics":{"box AP":"62.2","box APr":"55.1"},"uses_additional_data":true},{"leaderboard":"/sota/semantic-segmentation-on-ade20k","task":"Semantic Segmentation","dataset":"ADE20K","model":"EVA","rank_in_archive_order":6,"of":235,"metrics":{"Params (M)":"1074","Validation mIoU":"62.3"},"uses_additional_data":true},{"leaderboard":"/sota/semantic-segmentation-on-ade20k-val","task":"Semantic Segmentation","dataset":"ADE20K val","model":"EVA","rank_in_archive_order":3,"of":95,"metrics":{"mIoU":"61.5"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-coco-stuff-test","task":"Semantic Segmentation","dataset":"COCO-Stuff test","model":"EVA","rank_in_archive_order":3,"of":21,"metrics":{"mIoU":"53.4"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2211.07636","atlas_url":"https://app.syntology.ai/?focus=2211.07636","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2211.07636"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/rwightman/pytorch-image-models","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/PaddlePaddle/PASSL","reach":{"status":"ok","spdx":"Apache-2.0"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/leondgarse/keras_cv_attention_models/tree/main/keras_cv_attention_models/beit","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/baaivision/eva","reach":{"status":"ok","spdx":"MIT"}},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/PaddlePaddle/PaddleMIX/tree/develop/paddlemix/examples","reach":null},{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/open-mmlab/mmselfsup","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"ran_draft_wrong":1,"unverified":2},"by_repo_kind":{"official":{"samples":3,"ran":1,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"cad32ae74b9cae67","entry":"load_state_dict","repo":"baaivision/eva","repo_kind":"official","path":"EVA-01/clip/eva_clip.py","file_url":"https://github.com/baaivision/eva/blob/HEAD/EVA-01/clip/eva_clip.py","link_basis":"first_harvest_node","language":"python","status":"ran_draft_wrong","verification_level":1,"contract_check":"OUTPUT_MISDECLARED","metamorphic_tier":"deterministic","behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"cad32ae74b9cae67"}},{"code_sha256_prefix":"5c7cf820e99443ff","entry":"get_model_config","repo":"baaivision/eva","repo_kind":"official","path":"EVA-01/clip/eva_clip.py","file_url":"https://github.com/baaivision/eva/blob/HEAD/EVA-01/clip/eva_clip.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"5c7cf820e99443ff"}},{"code_sha256_prefix":"d503574dfff07e63","entry":"load_checkpoint","repo":"baaivision/eva","repo_kind":"official","path":"EVA-01/clip/eva_clip.py","file_url":"https://github.com/baaivision/eva/blob/HEAD/EVA-01/clip/eva_clip.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"d503574dfff07e63"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}