{"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/efficient-gender-classification-using-a-deep","title":"Efficient Gender Classification Using a Deep LDA-Pruned Net","arxiv_id":"1704.06305","date":"2017-04-20","proceeding":null,"authors":["Qing Tian","Tal Arbel","James J. Clark"],"abstract":"Many real-time tasks, such as human-computer interaction, require fast and\nefficient facial gender classification. Although deep CNN nets have been very\neffective for a multitude of classification tasks, their high space and time\ndemands make them impractical for personal computers and mobile devices without\na powerful GPU. In this paper, we develop a 16-layer, yet lightweight, neural\nnetwork which boosts efficiency while maintaining high accuracy. Our net is\npruned from the VGG-16 model starting from the last convolutional (conv) layer\nwhere we find neuron activations are highly uncorrelated given the gender.\nThrough Fisher's Linear Discriminant Analysis (LDA), we show that this high\ndecorrelation makes it safe to discard directly last conv layer neurons with\nhigh within-class variance and low between-class variance. Combined with either\nSupport Vector Machines (SVM) or Bayesian classification, the reduced CNNs are\ncapable of achieving comparable (or even higher) accuracies on the LFW and\nCelebA datasets than the original net with fully connected layers. On LFW, only\nfour Conv5_3 neurons are able to maintain a comparably high recognition\naccuracy, which results in a reduction of total network size by a factor of 70X\nwith a 11 fold speedup. Comparisons with a state-of-the-art pruning method as\nwell as two smaller nets in terms of accuracy loss and convolutional layers\npruning rate are also provided.","url_abs":"http://arxiv.org/abs/1704.06305v3","url_pdf":"http://arxiv.org/pdf/1704.06305v3.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":"efficient-gender-classification-using-a-deep","repo_url":"https://github.com/tchaton/deeplda","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"classification-1","task_name":"Classification"},{"task_slug":null,"task_name":"GPU"},{"task_slug":"gender-classification","task_name":"Gender Classification"},{"task_slug":"classification","task_name":"General Classification"}],"methods":[{"method_slug":"pruning","method_name":"Pruning"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}