{"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/dac-data-free-automatic-acceleration-of","title":"DAC: Data-free Automatic Acceleration of Convolutional Networks","arxiv_id":"1812.08374","date":"2018-12-20","proceeding":null,"authors":["Xin Li","Shuai Zhang","Bolan Jiang","Yingyong Qi","Mooi Choo Chuah","Ning Bi"],"abstract":"Deploying a deep learning model on mobile/IoT devices is a challenging task.\nThe difficulty lies in the trade-off between computation speed and accuracy. A\ncomplex deep learning model with high accuracy runs slowly on resource-limited\ndevices, while a light-weight model that runs much faster loses accuracy. In\nthis paper, we propose a novel decomposition method, namely DAC, that is\ncapable of factorizing an ordinary convolutional layer into two layers with\nmuch fewer parameters. DAC computes the corresponding weights for the newly\ngenerated layers directly from the weights of the original convolutional layer.\nThus, no training (or fine-tuning) or any data is needed. The experimental\nresults show that DAC reduces a large number of floating-point operations\n(FLOPs) while maintaining high accuracy of a pre-trained model. If 2% accuracy\ndrop is acceptable, DAC saves 53% FLOPs of VGG16 image classification model on\nImageNet dataset, 29% FLOPS of SSD300 object detection model on PASCAL VOC2007\ndataset, and 46% FLOPS of a multi-person pose estimation model on Microsoft\nCOCO dataset. Compared to other existing decomposition methods, DAC achieves\nbetter performance.","url_abs":"http://arxiv.org/abs/1812.08374v2","url_pdf":"http://arxiv.org/pdf/1812.08374v2.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":"dac-data-free-automatic-acceleration-of","repo_url":"https://github.com/baizhenmao95/2019-ZTE-Algorithm-Competition","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"caffe2","reach":null}],"tasks":[{"task_slug":"image-classification","task_name":"Image Classification"},{"task_slug":"multi-person-pose-estimation","task_name":"Multi-Person Pose Estimation"},{"task_slug":"object-detection","task_name":"Object Detection"},{"task_slug":"pose-estimation","task_name":"Pose Estimation"},{"task_slug":"image-classification","task_name":"image-classification"},{"task_slug":"object-detection-1","task_name":"object-detection"}],"methods":[{"method_slug":"speed","method_name":"SPEED"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}