{"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/billion-scale-semi-supervised-learning-for","title":"Billion-scale semi-supervised learning for image classification","arxiv_id":"1905.00546","date":"2019-05-02","proceeding":null,"authors":["I. Zeki Yalniz","Hervé Jégou","Kan Chen","Manohar Paluri","Dhruv Mahajan"],"abstract":"This paper presents a study of semi-supervised learning with large\nconvolutional networks. We propose a pipeline, based on a teacher/student\nparadigm, that leverages a large collection of unlabelled images (up to 1\nbillion). Our main goal is to improve the performance for a given target\narchitecture, like ResNet-50 or ResNext. We provide an extensive analysis of\nthe success factors of our approach, which leads us to formulate some\nrecommendations to produce high-accuracy models for image classification with\nsemi-supervised learning. As a result, our approach brings important gains to\nstandard architectures for image, video and fine-grained classification. For\ninstance, by leveraging one billion unlabelled images, our learned vanilla\nResNet-50 achieves 81.2% top-1 accuracy on the ImageNet benchmark.","url_abs":"http://arxiv.org/abs/1905.00546v1","url_pdf":"http://arxiv.org/pdf/1905.00546v1.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":"billion-scale-semi-supervised-learning-for","repo_url":"https://github.com/facebookresearch/semi-supervised-ImageNet1K-models","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"NOASSERTION"}},{"paper_slug":"billion-scale-semi-supervised-learning-for","repo_url":"https://github.com/leaderj1001/Billion-scale-semi-supervised-learning","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"billion-scale-semi-supervised-learning-for","repo_url":"https://github.com/salesforce/ensemble-of-averages","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"billion-scale-semi-supervised-learning-for","repo_url":"https://github.com/tiskw/patchcore-ad","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"object-recognition","task_name":"Object Recognition"},{"task_slug":"video-classification","task_name":"Video Classification"},{"task_slug":"image-classification","task_name":"image-classification"}],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"average-pooling","method_name":"Average Pooling"},{"method_slug":"batch-normalization","method_name":"Batch Normalization"},{"method_slug":"bottleneck-residual-block","method_name":"Bottleneck Residual Block"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"global-average-pooling","method_name":"Global Average Pooling"},{"method_slug":"grouped-convolution","method_name":"Grouped Convolution"},{"method_slug":"kaiming-initialization","method_name":"Kaiming Initialization"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"resnext","method_name":"ResNeXt"},{"method_slug":"resnext-block","method_name":"ResNeXt Block"},{"method_slug":"residual-block","method_name":"Residual Block"},{"method_slug":"residual-connection","method_name":"Residual Connection"},{"method_slug":"sgd-with-momentum","method_name":"SGD with Momentum"},{"method_slug":"weight-decay","method_name":"Weight Decay"}],"datasets_introduced":[{"slug":"ig-1b-targeted","name":"IG-1B-Targeted","full_name":"IG-1B-Targeted"}],"methods_introduced":[],"results":[{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ResNeXt-101 32x16d (semi-weakly sup.)","rank_in_archive_order":293,"of":1060,"metrics":{"Number of params":"193M","Top 1 Accuracy":"84.8%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ResNeXt-101 32x8d (semi-weakly sup.)","rank_in_archive_order":330,"of":1060,"metrics":{"Number of params":"88M","Top 1 Accuracy":"84.3%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ResNeXt-101 32x4d (semi-weakly sup.)","rank_in_archive_order":431,"of":1060,"metrics":{"Number of params":"42M","Top 1 Accuracy":"83.4%"},"uses_additional_data":false},{"leaderboard":"/sota/image-classification-on-omnibenchmark","task":"Image Classification","dataset":"OmniBenchmark","model":"IG-1B","rank_in_archive_order":6,"of":22,"metrics":{"Average Top-1 Accuracy":"40.4"},"uses_additional_data":true},{"leaderboard":"/sota/object-recognition-on-shape-bias","task":"Object Recognition","dataset":"shape bias","model":"SWSL (ResNeXt-101)","rank_in_archive_order":10,"of":18,"metrics":{"shape bias":"49.8"},"uses_additional_data":false},{"leaderboard":"/sota/object-recognition-on-shape-bias","task":"Object Recognition","dataset":"shape bias","model":"SWSL (ResNet-50)","rank_in_archive_order":16,"of":18,"metrics":{"shape bias":"28.6"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1905.00546","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}