Shiming Ge, Shengwei Zhao, Chenyu Li, Yu Zhang, Jia Li
{"title":"通过桥式蒸馏实现高效低分辨率人脸识别","authors":"Shiming Ge, Shengwei Zhao, Chenyu Li, Yu Zhang, Jia Li","doi":"arxiv-2409.11786","DOIUrl":null,"url":null,"abstract":"Face recognition in the wild is now advancing towards light-weight models,\nfast inference speed and resolution-adapted capability. In this paper, we\npropose a bridge distillation approach to turn a complex face model pretrained\non private high-resolution faces into a light-weight one for low-resolution\nface recognition. In our approach, such a cross-dataset resolution-adapted\nknowledge transfer problem is solved via two-step distillation. In the first\nstep, we conduct cross-dataset distillation to transfer the prior knowledge\nfrom private high-resolution faces to public high-resolution faces and generate\ncompact and discriminative features. In the second step, the resolution-adapted\ndistillation is conducted to further transfer the prior knowledge to synthetic\nlow-resolution faces via multi-task learning. By learning low-resolution face\nrepresentations and mimicking the adapted high-resolution knowledge, a\nlight-weight student model can be constructed with high efficiency and\npromising accuracy in recognizing low-resolution faces. Experimental results\nshow that the student model performs impressively in recognizing low-resolution\nfaces with only 0.21M parameters and 0.057MB memory. Meanwhile, its speed\nreaches up to 14,705, ~934 and 763 faces per second on GPU, CPU and mobile\nphone, respectively.","PeriodicalId":501480,"journal":{"name":"arXiv - CS - Multimedia","volume":"51 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Efficient Low-Resolution Face Recognition via Bridge Distillation\",\"authors\":\"Shiming Ge, Shengwei Zhao, Chenyu Li, Yu Zhang, Jia Li\",\"doi\":\"arxiv-2409.11786\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Face recognition in the wild is now advancing towards light-weight models,\\nfast inference speed and resolution-adapted capability. In this paper, we\\npropose a bridge distillation approach to turn a complex face model pretrained\\non private high-resolution faces into a light-weight one for low-resolution\\nface recognition. In our approach, such a cross-dataset resolution-adapted\\nknowledge transfer problem is solved via two-step distillation. In the first\\nstep, we conduct cross-dataset distillation to transfer the prior knowledge\\nfrom private high-resolution faces to public high-resolution faces and generate\\ncompact and discriminative features. In the second step, the resolution-adapted\\ndistillation is conducted to further transfer the prior knowledge to synthetic\\nlow-resolution faces via multi-task learning. By learning low-resolution face\\nrepresentations and mimicking the adapted high-resolution knowledge, a\\nlight-weight student model can be constructed with high efficiency and\\npromising accuracy in recognizing low-resolution faces. Experimental results\\nshow that the student model performs impressively in recognizing low-resolution\\nfaces with only 0.21M parameters and 0.057MB memory. Meanwhile, its speed\\nreaches up to 14,705, ~934 and 763 faces per second on GPU, CPU and mobile\\nphone, respectively.\",\"PeriodicalId\":501480,\"journal\":{\"name\":\"arXiv - CS - Multimedia\",\"volume\":\"51 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-18\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - CS - Multimedia\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.11786\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - CS - Multimedia","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.11786","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Efficient Low-Resolution Face Recognition via Bridge Distillation
Face recognition in the wild is now advancing towards light-weight models,
fast inference speed and resolution-adapted capability. In this paper, we
propose a bridge distillation approach to turn a complex face model pretrained
on private high-resolution faces into a light-weight one for low-resolution
face recognition. In our approach, such a cross-dataset resolution-adapted
knowledge transfer problem is solved via two-step distillation. In the first
step, we conduct cross-dataset distillation to transfer the prior knowledge
from private high-resolution faces to public high-resolution faces and generate
compact and discriminative features. In the second step, the resolution-adapted
distillation is conducted to further transfer the prior knowledge to synthetic
low-resolution faces via multi-task learning. By learning low-resolution face
representations and mimicking the adapted high-resolution knowledge, a
light-weight student model can be constructed with high efficiency and
promising accuracy in recognizing low-resolution faces. Experimental results
show that the student model performs impressively in recognizing low-resolution
faces with only 0.21M parameters and 0.057MB memory. Meanwhile, its speed
reaches up to 14,705, ~934 and 763 faces per second on GPU, CPU and mobile
phone, respectively.