{"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/deit-iii-revenge-of-the-vit","title":"DeiT III: Revenge of the ViT","arxiv_id":"2204.07118","date":"2022-04-14","proceeding":null,"authors":["Hugo Touvron","Matthieu Cord","Hervé Jégou"],"abstract":"A Vision Transformer (ViT) is a simple neural architecture amenable to serve several computer vision tasks. It has limited built-in architectural priors, in contrast to more recent architectures that incorporate priors either about the input data or of specific tasks. Recent works show that ViTs benefit from self-supervised pre-training, in particular BerT-like pre-training like BeiT. In this paper, we revisit the supervised training of ViTs. Our procedure builds upon and simplifies a recipe introduced for training ResNet-50. It includes a new simple data-augmentation procedure with only 3 augmentations, closer to the practice in self-supervised learning. Our evaluations on Image classification (ImageNet-1k with and without pre-training on ImageNet-21k), transfer learning and semantic segmentation show that our procedure outperforms by a large margin previous fully supervised training recipes for ViT. It also reveals that the performance of our ViT trained with supervision is comparable to that of more recent architectures. Our results could serve as better baselines for recent self-supervised approaches demonstrated on ViT.","url_abs":"https://arxiv.org/abs/2204.07118v1","url_pdf":"https://arxiv.org/pdf/2204.07118v1.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":"deit-iii-revenge-of-the-vit","repo_url":"https://github.com/facebookresearch/deit","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"deit-iii-revenge-of-the-vit","repo_url":"https://github.com/affjljoo3581/deit3-jax","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"jax","reach":null},{"paper_slug":"deit-iii-revenge-of-the-vit","repo_url":"https://github.com/rwightman/pytorch-image-models","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"deit-iii-revenge-of-the-vit","repo_url":"https://github.com/2023-MindSpore-1/ms-code-7/tree/main/vit_base","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"deit-iii-revenge-of-the-vit","repo_url":"https://github.com/2024-MindSpore-1/Code2/tree/main/model-1/deit","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"deit-iii-revenge-of-the-vit","repo_url":"https://github.com/Mind23-2/MindCode-121","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":{"status":"ok"}},{"paper_slug":"deit-iii-revenge-of-the-vit","repo_url":"https://github.com/Mind23-2/MindCode-2/tree/main/vit_base","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"deit-iii-revenge-of-the-vit","repo_url":"https://github.com/MindSpore-paper-code-2/code3/tree/main/vit_base","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"deit-iii-revenge-of-the-vit","repo_url":"https://github.com/alibaba/EasyCV","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null},{"paper_slug":"deit-iii-revenge-of-the-vit","repo_url":"https://github.com/open-mmlab/mmclassification","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"Apache-2.0"}},{"paper_slug":"deit-iii-revenge-of-the-vit","repo_url":"https://github.com/yangyucheng000/MSpaper/tree/main/vit_base","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"deit-iii-revenge-of-the-vit","repo_url":"https://gitlab.com/birder/birder","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"data-augmentation","task_name":"Data Augmentation"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"self-supervised-learning","task_name":"Self-Supervised Learning"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[{"method_slug":"3-augment","method_name":"3-Augment"},{"method_slug":"absolute-position-encodings","method_name":"Absolute Position Encodings"},{"method_slug":"adam","method_name":"Adam"},{"method_slug":"attention","method_name":"Attention"},{"method_slug":"bpe","method_name":"BPE"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"fixres","method_name":"FixRes"},{"method_slug":"label-smoothing","method_name":"Label Smoothing"},{"method_slug":"layer-normalization","method_name":"Layer Normalization"},{"method_slug":"layerscale","method_name":"LayerScale"},{"method_slug":"linear-layer","method_name":"Linear Layer"},{"method_slug":"multi-head-attention","method_name":"Multi-Head Attention"},{"method_slug":"position-wise-feed-forward-layer","method_name":"Position-Wise Feed-Forward Layer"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"transformer","method_name":"Transformer"},{"method_slug":"vision-transformer","method_name":"Vision Transformer"}],"datasets_introduced":[],"methods_introduced":[{"slug":"3-augment","name":"3-Augment","full_name":"3-Augment"}],"results":[{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ViT-B @384 (DeiT III, 21k)","rank_in_archive_order":126,"of":1060,"metrics":{"Top 1 Accuracy":"86.7%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ViT-L","rank_in_archive_order":198,"of":1060,"metrics":{"GFLOPs":"191.2","Number of params":"304.8M","Top 1 Accuracy":"85.8%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ViT-B @224 (DeiT III, 21k)","rank_in_archive_order":207,"of":1060,"metrics":{"Top 1 Accuracy":"85.7%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ViT-H @224 (DeiT III)","rank_in_archive_order":245,"of":1060,"metrics":{"Top 1 Accuracy":"85.2%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ViT-B @384 (DeiT III)","rank_in_archive_order":270,"of":1060,"metrics":{"Number of params":"87M","Top 1 Accuracy":"85.0%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ViT-L @224 (DeiT III)","rank_in_archive_order":274,"of":1060,"metrics":{"Top 1 Accuracy":"84.9%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ViT-B @224 (DeiT III)","rank_in_archive_order":386,"of":1060,"metrics":{"Top 1 Accuracy":"83.8%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ViT-S @384 (DeiT III)","rank_in_archive_order":428,"of":1060,"metrics":{"GFLOPs":"15.5","Number of params":"22M","Top 1 Accuracy":"83.4%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ViT-S @224 (DeiT III, 21k)","rank_in_archive_order":464,"of":1060,"metrics":{"Top 1 Accuracy":"83.1%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ViT-S @224 (DeiT III)","rank_in_archive_order":643,"of":1060,"metrics":{"Top 1 Accuracy":"81.4%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet-real","task":"Image Classification","dataset":"ImageNet ReaL","model":"ViT-L @384 (DeiT III, 21k)","rank_in_archive_order":54,"of":57,"metrics":{"Number of params":"304M","Top 1 Accuracy":"87.7%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet-real","task":"Image Classification","dataset":"ImageNet ReaL","model":"ViT-H @224 (DeiT III, 21k)","rank_in_archive_order":55,"of":57,"metrics":{"Number of params":"632M","Top 1 Accuracy":"87.2%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet-real","task":"Image Classification","dataset":"ImageNet ReaL","model":"ViT-L @224 (DeiT III, 21k)","rank_in_archive_order":56,"of":57,"metrics":{"Top 1 Accuracy":"87.0%"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-ade20k-val","task":"Semantic Segmentation","dataset":"ADE20K val","model":"DeiT-L","rank_in_archive_order":29,"of":95,"metrics":{"mIoU":"55.6"},"uses_additional_data":false},{"leaderboard":"/sota/semantic-segmentation-on-ade20k-val","task":"Semantic Segmentation","dataset":"ADE20K val","model":"DeiT-B","rank_in_archive_order":36,"of":95,"metrics":{"mIoU":"54.1"},"uses_additional_data":false}],"syntology":{"syntology_url":"https://syntology.ai/paper/2204.07118","atlas_url":"https://app.syntology.ai/?focus=2204.07118","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}