takuya-takeuchi / facerecognitiondotnet

The world's simplest facial recognition api for .NET on Windows, MacOS and Linux

The world's simplest facial recognition api for .NET
This repository is porting https://github.com/ageitgey/face_recognition by C#.

This package supports cross platform, Windows, Linux and MacOSX!!

Package	OS	x86	x64	ARM	ARM64	Nuget
FaceRecognitionDotNet (CPU)	Windows	-	✓	-	-
	Linux	-	✓	-	-
	OSX	-	✓	-	-
FaceRecognitionDotNet for CUDA 9.2	Windows	-	✓	-	-
	Linux	-	✓	-	-
	OSX	-	-	-	-
FaceRecognitionDotNet for CUDA 10.0	Windows	-	✓	-	-
	Linux	-	✓	-	-
	OSX	-	-	-	-
FaceRecognitionDotNet for CUDA 10.1	Windows	-	✓	-	-
	Linux	-	✓	-	-
	OSX	-	-	-	-
FaceRecognitionDotNet for CUDA 10.2	Windows	-	✓	-	-
	Linux	-	✓	-	-
	OSX	-	-	-	-
FaceRecognitionDotNet for CUDA 11.0	Windows	-	✓	-	-
	Linux	-	✓	-	-
	OSX	-	-	-	-
FaceRecognitionDotNet for CUDA 11.1	Windows	-	✓	-	-
	Linux	-	✓	-	-
	OSX	-	-	-	-
FaceRecognitionDotNet for CUDA 11.2	Windows	-	✓	-	-
	Linux	-	✓	-	-
	OSX	-	-	-	-
FaceRecognitionDotNet for Intel MKL	Windows	-	✓	-	-
	Linux	-	✓	-	-
	OSX	-	✓	-	-
FaceRecognitionDotNet for ARM	Windows	-	-	-	-
	Linux	-	-	-	-
	OSX	-	-	-	-

⚠️ FaceRecognitionDotNet for ARM is not tested yet

Support API

face_recognition API	Corresponding API	Note
batch_face_locations	BatchFaceLocations
compare_faces	CompareFaces
face_distance	FaceDistance
face_encodings	FaceEncodings
face_landmarks	FaceLandmarks	And support Helen dataset ⚠️
face_locations	FaceLocations	And support to get confidence and use custom face detector
load_image_file	LoadImageFile
-	CropFaces	Crop image with specified locations
-	EyeBlinkDetect	Detect person is blinking or not Support Large model and Helen dataset ⚠️
-	LoadImage	From memory data or System.Drawing.Bitmap
-	PredictAge	Predict human age. Use Adience Benchmark Of Unfiltered Faces For Gender And Age Classification dataset ⚠️
-	PredictEmotion	Predict emotion for human face. Use Corrective re-annotation of FER - CK+ - KDEF ⚠️
-	PredictGender	Predict human gender. Use UTKFace dataset ⚠️
-	PredictProbabilityAge	Predict probability of human age. Use Adience Benchmark Of Unfiltered Faces For Gender And Age Classification dataset ⚠️
-	PredictProbabilityEmotion	Predict probability of emotion from human face. Use Corrective re-annotation of FER - CK+ - KDEF ⚠️
-	PredictProbabilityGender	Predict probability of human gender. Use UTKFace dataset ⚠️
-	PredictHeadPose	Predict human head pose. Use 300W-LP dataset ⚠️

⚠️ Warning

You must train dataset by yourself. I will NOT provide pretrained model file due to avoiding license issue. You can check the following examples to train dataset.

• tools/AgeTraining
• tools/EmotionTraining
• tools/EmotionTrainingV2
• tools/GenderTraining
• tools/HeadPoseTraining
• tools/HelenTraining

Demo

Face Recognition

Other Face Functions

Face Landmark	Age and Gender Classification	Head Pose Estimation	Emotion Estimation

Document

FaceRecognitionDotNet support full xml document for Visual Studio. A xml document is written English and Japanese. And you can check online document at FaceRecognitionDotNet API Document

Dependencies Libraries and Products

face_recognition

License: The MIT License

Author: Adam Geitgey

Principal Use: The world's simplest facial recognition api for Python and the command line. Main goal of FaceRecognitionDotNet is what ports face_recognition by C#.

face_recognition_models

License: Creative Commons Zero v1.0 Universal License

Author: Adam Geitgey

Principal Use: Trained models for the face_recognition python library

dlib

License: Boost Software License

Author: Davis E. King

Principal Use: A toolkit for making real world machine learning and data analysis applications in C++.

DlibDotNet

License: The MIT License

Author: Takuya Takeuchi

Principal Use: Use dlib interface via .NET. This library is developed by this owner.

OpenCVSharp

License: The BSD 3-Clause License

Author: shimat

Principal Use: Loading image data by opencv wrapper for example