{"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/lsun-construction-of-a-large-scale-image","title":"LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop","arxiv_id":"1506.03365","date":"2015-06-10","proceeding":null,"authors":["Fisher Yu","Ari Seff","yinda zhang","Shuran Song","Thomas Funkhouser","Jianxiong Xiao"],"abstract":"While there has been remarkable progress in the performance of visual\nrecognition algorithms, the state-of-the-art models tend to be exceptionally\ndata-hungry. Large labeled training datasets, expensive and tedious to produce,\nare required to optimize millions of parameters in deep network models. Lagging\nbehind the growth in model capacity, the available datasets are quickly\nbecoming outdated in terms of size and density. To circumvent this bottleneck,\nwe propose to amplify human effort through a partially automated labeling\nscheme, leveraging deep learning with humans in the loop. Starting from a large\nset of candidate images for each category, we iteratively sample a subset, ask\npeople to label them, classify the others with a trained model, split the set\ninto positives, negatives, and unlabeled based on the classification\nconfidence, and then iterate with the unlabeled set. To assess the\neffectiveness of this cascading procedure and enable further progress in visual\nrecognition research, we construct a new image dataset, LSUN. It contains\naround one million labeled images for each of 10 scene categories and 20 object\ncategories. We experiment with training popular convolutional networks and find\nthat they achieve substantial performance gains when trained on this dataset.","url_abs":"http://arxiv.org/abs/1506.03365v3","url_pdf":"http://arxiv.org/pdf/1506.03365v3.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":"lsun-construction-of-a-large-scale-image","repo_url":"https://github.com/fyu/lsun","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"none","reach":{"status":"ok"}},{"paper_slug":"lsun-construction-of-a-large-scale-image","repo_url":"https://github.com/monniert/unicorn","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"lsun-construction-of-a-large-scale-image","repo_url":"https://github.com/psanch21/imp_bigan","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"lsun-construction-of-a-large-scale-image","repo_url":"https://github.com/swing-research/deepmesh","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[{"slug":"lsun","name":"LSUN","full_name":"Large-scale Scene UNderstanding Challenge"}],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1506.03365","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}