人脸属性分析--性别、年龄和表情识别

人脸属性指的是根据给定的人脸判断其性别、年龄和表情等，当前在github上开源了一些相关的工作，大部分都是基于tensorflow的，还有一部分是keras，CVPR2015曾有一篇是用caffe做的.

CSDN

从0到1实现基于Tornado和Tensorflow的人脸、年龄、性别识别

基于caffe的表情识别

tensorflow练习12：利用图片预测年龄与性别

怎样用Keras识别人物面部表情

github

https://github.com/GilLevi/AgeGenderDeepLearning:CVPR2015 caffe实现

https://github.com/dpressel/rude-carnie：CVPR2015对应的tensorflow实现

https://github.com/truongnmt/multi-task-learning: DEX: Deep EXpectation 实现

https://github.com/ZZUTK/Face-Aging-CAAE:CVPR2017 Age Progression/Regression by Conditional Adversarial Autoencoder

https://github.com/BoyuanJiang/Age-Gender-Estimate-TF:使用inception v1同时预测性别和年龄，受限于使用的dlib检测器，效果并不是很好

https://github.com/zZyan/race_gender_recognition:gender Accuracy: 0.951493,race Accuracy: 0.87557212

https://github.com/yu4u/age-gender-estimation:UTKFace库 WideResNet,64*64输入 keras模型，MAE 4.06

https://github.com/dandynaufaldi/Agendernetkeras实现了inception v3,mobilenet和ssr在imdb、utkface等的训练

https://github.com/jocialiang/gender_classifier:性别识别全流程实现 94% accuracy

https://github.com/oarriaga/face_classification:表情识别

https://github.com/wondonghyeon/face-classification:性别和种族识别

https://github.com/shamangary/SSR-Net:年龄识别

https://github.com/b02901145/SSR-Net_megaage-asian：亚洲人优化

https://github.com/yu4u/age-gender-estimation：年龄和性别识别

https://github.com/isseu/emotion-recognition-neural-networks:表情66% with fer2013，性别96% with imdb.

https://github.com/zealerww/gender_age_classification:91% accuracy in gender and 55% in age

https://github.com/vipstone/faceai:gender 96%

https://github.com/XiuweiHe/EmotionClassifier:表情识别

https://github.com/HectorAnadon/Face-expression-and-ethnic-recognition:表情和种族识别

https://github.com/ybch14/Facial-Expression-Recognition-ResNet66.7% on fer2013 with resnet50

https://github.com/JostineHo/mememoji:58% with 动画展示

https://github.com/mangorocoro/racedetector:种族识别

https://github.com/HectorAnadon/Face-expression-and-ethnic-recognition:表情 72% accuracy ，种族95% accuracy

https://github.com/XiuweiHe/EmotionClassifier： 66% on fer2013 with mini_XCEPTION

https://github.com/truongnmt/multi-task-learning:多任务学习

useless

https://github.com/StevenKe8080/recognition_gender:使用爬取的图片训练

https://github.com/zonetrooper32/AgeEstimateAdience:

https://github.com/OValery16/gender-age-classification:

数据库

UTKFace：over 20,000 face images with annotations of age, gender, and ethnicity. The images cover large variation in pose, facial expression, illumination, occlusion, resolution, etc.

SCUT-FBP5500:5500 frontal faces with diverse properties (male/female, Asian/Caucasian, ages) and diverse labels (facial landmarks, beauty scores in 5 scales, beauty score distribution), which allows different computational model with different facial beauty prediction paradigms, such as appearance-based/shape-based facial beauty classification/regression/ranking model for male/female of Asian/Caucasian

CelebA:标注了40个属性，第21个属性为性别

202,599 number of face images, and
5 landmark locations, 40 binary attributes annotations per image.

APPA-REAL ：视觉年龄估计，7,591张带有实际年龄和视觉年龄标注的图片，分为 4113 train, 1500 valid and 1978 test images,大小:844M

AFAD Dataset： Asian Face Age Dataset，more than 160K facial images and the corresponding age and gender labels.暂未开放下载

FER+ :微软重新标注的fer2013，表情识别比赛数据

NKI:GENKI数据集是由加利福尼亚大学的机器概念实验室收集。该数据集包含GENKI-R2009a,GENKI-4K,GENKI-SZSL三个部分。GENKI-R2009a包含11159个图像，GENKI-4K包含4000个图像，分为“笑”和“不笑”两种，每个图片的人脸的尺度大小，姿势，光照变化，头的转动等都不一样，专门用于做笑脸识别。GENKI-SZSL包含3500个图像，这些图像包括广泛的背景，光照条件，地理位置，个人身份和种族等

Datasets	Description	Links	Key features	Publish Time
CelebA	10,177 number of identities, 202,599 number of face images, and 5 landmark locations, 40 binary attributes annotations per image.	Download	attribute & landmark	2015
IMDB-WIKI	500k+ face images with age and gender labels	Download	age & gender	2015
Adience	Unfiltered faces for gender and age classification	Download	age & gender	2014
WFLW?	WFLW contains 10000 faces (7500 for training and 2500 for testing) with 98 fully manual annotated landmarks.	Download	landmarks	2018
Caltech10k Web Faces	The dataset has 10,524 human faces of various resolutions and in different settings	Download	landmarks	2005
EmotioNet	The EmotioNet database includes950,000 images with annotated AUs. A subset of the images in the EmotioNet database correspond to basic and compound emotions.	Download	AU and Emotion	2017
RAF( Real-world Affective Faces)	29672 number of real-world images, including 7 classes of basic emotions and 12 classes of compound emotions, 5 accurate landmark locations, 37 automatic landmark locations, race, age range and gender attributes annotations per image	Download	Emotions、landmark、race、age and gender	2017