Skip to content

适用于移动端的人脸识别模型,计算量与mobilefacenet相同,但megaface上提升了2%+

Notifications You must be signed in to change notification settings

onexuan/facerecognize-for-mobile-phone

 
 

Repository files navigation

facerecognize-for-mobile-phone

适用于移动端的人脸识别模型,计算量于mobilefacenet相同,但megaface上提升了2%+。我训练的mobilefacenet在megaface VER上比原作者 高出了2%,因为训练的数据和方法不一样。

模型在各个数据集上表现如下:

Methods Flops (112x112) LFW CFP-FP AgeDB Megaface-Id Megaface-Ver@1e-6 备 注
MobileFaceNet440,R 440M 99.70+ 96.70+ 96.95+ 92.85+ 94.20+ 未开源
ZW350 356M 99.70+ 96.82+ 97.00+ 93.90+ 94.70+ 未开源
ZW400 404M 99.70+ 96.95+ 97.00+ 94.46+ 95.60+ 未开源
MobileFaceNet600,R 612M 99.76+ 97.60+ 97.50+ 95.14+ 95.98+ 已开源
ZW440 444M 99.76+ 97.30+ 97.40+ 95.25+ 96.00+ 已开源

Megafce测试结果图

zw440-id zw440-id

zw440-ver

zw440-ver

速度比对测试

设备:i5-6500(暂时没有移动设备测试)

Methods Openvino opencv单线程
MobileFaceNet600,R 6ms 141ms
ZW440 7ms 80ms

移动设备

经过测试,zw440并没有Mobilefacenet600M快.感谢moli的测试

模型地址

模型包含mxnet ncnn caffe 三种格式 Baidu Drive 提取码:b0dm

训练数据

https://github.com/deepinsight/insightface/tree/master/iccv19-challenge

参考项目

https://github.com/deepinsight/insightface

https://github.com/happynear/FaceVerification

https://github.com/Tencent/ncnn

https://github.com/cypw/MXNet2Caffe

Todo

没有做速度方面考虑,后期跟进改善。

About

适用于移动端的人脸识别模型,计算量与mobilefacenet相同,但megaface上提升了2%+

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published