{"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/design-of-efficient-convolutional-layers","title":"Design of Efficient Convolutional Layers using Single Intra-channel Convolution, Topological Subdivisioning and Spatial \"Bottleneck\" Structure","arxiv_id":"1608.04337","date":"2016-08-15","proceeding":null,"authors":["Min Wang","Baoyuan Liu","Hassan Foroosh"],"abstract":"Deep convolutional neural networks achieve remarkable visual recognition\nperformance, at the cost of high computational complexity. In this paper, we\nhave a new design of efficient convolutional layers based on three schemes. The\n3D convolution operation in a convolutional layer can be considered as\nperforming spatial convolution in each channel and linear projection across\nchannels simultaneously. By unravelling them and arranging the spatial\nconvolution sequentially, the proposed layer is composed of a single\nintra-channel convolution, of which the computation is negligible, and a linear\nchannel projection. A topological subdivisioning is adopted to reduce the\nconnection between the input channels and output channels. Additionally, we\nalso introduce a spatial \"bottleneck\" structure that utilizes a\nconvolution-projection-deconvolution pipeline to take advantage of the\ncorrelation between adjacent pixels in the input. Our experiments demonstrate\nthat the proposed layers remarkably outperform the standard convolutional\nlayers with regard to accuracy/complexity ratio. Our models achieve similar\naccuracy to VGG, ResNet-50, ResNet-101 while requiring 42, 4.5, 6.5 times less\ncomputation respectively.","url_abs":"http://arxiv.org/abs/1608.04337v2","url_pdf":"http://arxiv.org/pdf/1608.04337v2.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":"design-of-efficient-convolutional-layers","repo_url":"https://github.com/asfathermou/human-computer-interaction","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[],"methods":[{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1608.04337","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}