{"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/dropblock-a-regularization-method-for","title":"DropBlock: A regularization method for convolutional networks","arxiv_id":"1810.12890","date":"2018-10-30","proceeding":"NeurIPS 2018 12","authors":["Golnaz Ghiasi","Tsung-Yi Lin","Quoc V. Le"],"abstract":"Deep neural networks often work well when they are over-parameterized and\ntrained with a massive amount of noise and regularization, such as weight decay\nand dropout. Although dropout is widely used as a regularization technique for\nfully connected layers, it is often less effective for convolutional layers.\nThis lack of success of dropout for convolutional layers is perhaps due to the\nfact that activation units in convolutional layers are spatially correlated so\ninformation can still flow through convolutional networks despite dropout. Thus\na structured form of dropout is needed to regularize convolutional networks. In\nthis paper, we introduce DropBlock, a form of structured dropout, where units\nin a contiguous region of a feature map are dropped together. We found that\napplying DropbBlock in skip connections in addition to the convolution layers\nincreases the accuracy. Also, gradually increasing number of dropped units\nduring training leads to better accuracy and more robust to hyperparameter\nchoices. Extensive experiments show that DropBlock works better than dropout in\nregularizing convolutional networks. On ImageNet classification, ResNet-50\narchitecture with DropBlock achieves $78.13\\%$ accuracy, which is more than\n$1.6\\%$ improvement on the baseline. On COCO detection, DropBlock improves\nAverage Precision of RetinaNet from $36.8\\%$ to $38.4\\%$.","url_abs":"http://arxiv.org/abs/1810.12890v1","url_pdf":"http://arxiv.org/pdf/1810.12890v1.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":"dropblock-a-regularization-method-for","repo_url":"https://github.com/DHZS/tf-dropblock","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"dropblock-a-regularization-method-for","repo_url":"https://github.com/gan3sh500/dropblock","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}},{"paper_slug":"dropblock-a-regularization-method-for","repo_url":"https://github.com/iantimmis/DropBlock-Keras-Implementation","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"jax","reach":{"status":"ok"}},{"paper_slug":"dropblock-a-regularization-method-for","repo_url":"https://github.com/miguelvr/dropblock","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"dropblock-a-regularization-method-for","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":"dropblock-a-regularization-method-for","repo_url":"https://github.com/yuyijie1995/dropblock_mxnet_bottom_implemention","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"mxnet","reach":{"status":"ok"}},{"paper_slug":"dropblock-a-regularization-method-for","repo_url":"https://github.com/MindCode-4/code-11/tree/main/dropblock-a-regularization-method","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"dropblock-a-regularization-method-for","repo_url":"https://github.com/MindCode-4/code-6/tree/main/dropblock-a-regularization-method","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"dropblock-a-regularization-method-for","repo_url":"https://github.com/mindspore-ai/contrib/tree/master/application/dropblock-a-regularization-method-for-convolutional-networks","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null},{"paper_slug":"dropblock-a-regularization-method-for","repo_url":"https://github.com/nanzhaogang/contrib/tree/master/application/dropblock-a-regularization-method-for-convolutional-networks","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"mindspore","reach":null}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"object-detection","task_name":"Object Detection"}],"methods":[{"method_slug":"1x1-convolution","method_name":"1x1 Convolution"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dropblock","method_name":"DropBlock"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"fpn","method_name":"FPN"},{"method_slug":"focal-loss","method_name":"Focal Loss"},{"method_slug":"retinanet","method_name":"RetinaNet"}],"datasets_introduced":[],"methods_introduced":[{"slug":"dropblock","name":"DropBlock","full_name":"DropBlock"}],"results":[{"leaderboard":"/sota/image-classification-on-imagenet","task":"Image Classification","dataset":"ImageNet","model":"ResNet-50 + DropBlock (0.9 kp, 0.1 label smoothing)","rank_in_archive_order":840,"of":1060,"metrics":{"Top 1 Accuracy":"78.35%"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1810.12890","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}