{"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/channelnets-compact-and-efficient","title":"ChannelNets: Compact and Efficient Convolutional Neural Networks via Channel-Wise Convolutions","arxiv_id":"1809.01330","date":"2018-09-05","proceeding":"NeurIPS 2018 12","authors":["Hongyang Gao","Zhengyang Wang","Shuiwang Ji"],"abstract":"Convolutional neural networks (CNNs) have shown great capability of solving\nvarious artificial intelligence tasks. However, the increasing model size has\nraised challenges in employing them in resource-limited applications. In this\nwork, we propose to compress deep models by using channel-wise convolutions,\nwhich re- place dense connections among feature maps with sparse ones in CNNs.\nBased on this novel operation, we build light-weight CNNs known as ChannelNets.\nChannel- Nets use three instances of channel-wise convolutions; namely group\nchannel-wise convolutions, depth-wise separable channel-wise convolutions, and\nthe convolu- tional classification layer. Compared to prior CNNs designed for\nmobile devices, ChannelNets achieve a significant reduction in terms of the\nnumber of parameters and computational cost without loss in accuracy. Notably,\nour work represents the first attempt to compress the fully-connected\nclassification layer, which usually accounts for about 25% of total parameters\nin compact CNNs. Experimental results on the ImageNet dataset demonstrate that\nChannelNets achieve consistently better performance compared to prior methods.","url_abs":"http://arxiv.org/abs/1809.01330v1","url_pdf":"http://arxiv.org/pdf/1809.01330v1.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":"channelnets-compact-and-efficient","repo_url":"https://github.com/HongyangGao/ChannelNets","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"channelnets-compact-and-efficient","repo_url":"https://github.com/osmr/imgclsmob","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"mxnet","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"classification","task_name":"General Classification"}],"methods":[{"method_slug":"dense-connections","method_name":"Dense Connections"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}