{"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/blockdrop-dynamic-inference-paths-in-residual","title":"BlockDrop: Dynamic Inference Paths in Residual Networks","arxiv_id":"1711.08393","date":"2017-11-22","proceeding":"CVPR 2018 6","authors":["Zuxuan Wu","Tushar Nagarajan","Abhishek Kumar","Steven Rennie","Larry S. Davis","Kristen Grauman","Rogerio Feris"],"abstract":"Very deep convolutional neural networks offer excellent recognition results,\nyet their computational expense limits their impact for many real-world\napplications. We introduce BlockDrop, an approach that learns to dynamically\nchoose which layers of a deep network to execute during inference so as to best\nreduce total computation without degrading prediction accuracy. Exploiting the\nrobustness of Residual Networks (ResNets) to layer dropping, our framework\nselects on-the-fly which residual blocks to evaluate for a given novel image.\nIn particular, given a pretrained ResNet, we train a policy network in an\nassociative reinforcement learning setting for the dual reward of utilizing a\nminimal number of blocks while preserving recognition accuracy. We conduct\nextensive experiments on CIFAR and ImageNet. The results provide strong\nquantitative and qualitative evidence that these learned policies not only\naccelerate inference but also encode meaningful visual information. Built upon\na ResNet-101 model, our method achieves a speedup of 20\\% on average, going as\nhigh as 36\\% for some images, while maintaining the same 76.4\\% top-1 accuracy\non ImageNet.","url_abs":"http://arxiv.org/abs/1711.08393v4","url_pdf":"http://arxiv.org/pdf/1711.08393v4.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":"blockdrop-dynamic-inference-paths-in-residual","repo_url":"https://github.com/Tushar-N/blockdrop","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"}],"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":"kaiming-initialization","method_name":"Kaiming Initialization"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"residual-block","method_name":"Residual Block"},{"method_slug":"residual-connection","method_name":"Residual Connection"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1711.08393","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}