{"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/chamnet-towards-efficient-network-design","title":"ChamNet: Towards Efficient Network Design through Platform-Aware Model Adaptation","arxiv_id":"1812.08934","date":"2018-12-21","proceeding":"CVPR 2019 6","authors":["Xiaoliang Dai","Peizhao Zhang","Bichen Wu","Hongxu Yin","Fei Sun","Yanghan Wang","Marat Dukhan","Yunqing Hu","Yiming Wu","Yangqing Jia","Peter Vajda","Matt Uyttendaele","Niraj K. Jha"],"abstract":"This paper proposes an efficient neural network (NN) architecture design\nmethodology called Chameleon that honors given resource constraints. Instead of\ndeveloping new building blocks or using computationally-intensive reinforcement\nlearning algorithms, our approach leverages existing efficient network building\nblocks and focuses on exploiting hardware traits and adapting computation\nresources to fit target latency and/or energy constraints. We formulate\nplatform-aware NN architecture search in an optimization framework and propose\na novel algorithm to search for optimal architectures aided by efficient\naccuracy and resource (latency and/or energy) predictors. At the core of our\nalgorithm lies an accuracy predictor built atop Gaussian Process with Bayesian\noptimization for iterative sampling. With a one-time building cost for the\npredictors, our algorithm produces state-of-the-art model architectures on\ndifferent platforms under given constraints in just minutes. Our results show\nthat adapting computation resources to building blocks is critical to model\nperformance. Without the addition of any bells and whistles, our models achieve\nsignificant accuracy improvements against state-of-the-art hand-crafted and\nautomatically designed architectures. We achieve 73.8% and 75.3% top-1 accuracy\non ImageNet at 20ms latency on a mobile CPU and DSP. At reduced latency, our\nmodels achieve up to 8.5% (4.8%) and 6.6% (9.3%) absolute top-1 accuracy\nimprovements compared to MobileNetV2 and MnasNet, respectively, on a mobile CPU\n(DSP), and 2.7% (4.6%) and 5.6% (2.6%) accuracy gains over ResNet-101 and\nResNet-152, respectively, on an Nvidia GPU (Intel CPU).","url_abs":"http://arxiv.org/abs/1812.08934v1","url_pdf":"http://arxiv.org/pdf/1812.08934v1.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":"chamnet-towards-efficient-network-design","repo_url":"https://github.com/facebookresearch/mobile-vision","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"caffe2","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[{"task_slug":"bayesian-optimization","task_name":"Bayesian Optimization"},{"task_slug":null,"task_name":"CPU"},{"task_slug":"efficient-neural-network","task_name":"Efficient Neural Network"},{"task_slug":null,"task_name":"GPU"},{"task_slug":"architecture-search","task_name":"Neural Architecture Search"},{"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":"convolution","method_name":"Convolution"},{"method_slug":"dense-connections","method_name":"Dense Connections"},{"method_slug":"depthwise-convolution","method_name":"Depthwise Convolution"},{"method_slug":"depthwise-separable-convolution","method_name":"Depthwise Separable Convolution"},{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"gaussian-process","method_name":"Gaussian Process"},{"method_slug":"global-average-pooling","method_name":"Global Average Pooling"},{"method_slug":"inverted-residual-block","method_name":"Inverted Residual Block"},{"method_slug":"mnasnet","method_name":"MnasNet"},{"method_slug":"pointwise-convolution","method_name":"Pointwise Convolution"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"sigmoid-activation","method_name":"Sigmoid Activation"},{"method_slug":"softmax","method_name":"Softmax"},{"method_slug":"squeeze-and-excitation-block","method_name":"Squeeze-and-Excitation Block"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1812.08934","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}