{"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/sbnet-sparse-blocks-network-for-fast","title":"SBNet: Sparse Blocks Network for Fast Inference","arxiv_id":"1801.02108","date":"2018-01-07","proceeding":"CVPR 2018 6","authors":["Mengye Ren","Andrei Pokrovsky","Bin Yang","Raquel Urtasun"],"abstract":"Conventional deep convolutional neural networks (CNNs) apply convolution\noperators uniformly in space across all feature maps for hundreds of layers -\nthis incurs a high computational cost for real-time applications. For many\nproblems such as object detection and semantic segmentation, we are able to\nobtain a low-cost computation mask, either from a priori problem knowledge, or\nfrom a low-resolution segmentation network. We show that such computation masks\ncan be used to reduce computation in the high-resolution main network. Variants\nof sparse activation CNNs have previously been explored on small-scale tasks\nand showed no degradation in terms of object classification accuracy, but often\nmeasured gains in terms of theoretical FLOPs without realizing a practical\nspeed-up when compared to highly optimized dense convolution implementations.\nIn this work, we leverage the sparsity structure of computation masks and\npropose a novel tiling-based sparse convolution algorithm. We verified the\neffectiveness of our sparse CNN on LiDAR-based 3D object detection, and we\nreport significant wall-clock speed-ups compared to dense convolution without\nnoticeable loss of accuracy.","url_abs":"http://arxiv.org/abs/1801.02108v2","url_pdf":"http://arxiv.org/pdf/1801.02108v2.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":"sbnet-sparse-blocks-network-for-fast","repo_url":"https://github.com/uber/sbnet","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":{"status":"unanswered"}},{"paper_slug":"sbnet-sparse-blocks-network-for-fast","repo_url":"https://github.com/Omar-Awad/SCNN_GPU2","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"3d-object-detection","task_name":"3D Object Detection"},{"task_slug":"object","task_name":"Object"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"convolution","method_name":"Convolution"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1801.02108","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}