Face Recognition API

API built using flask returns face embedding vector as response.

Getting Started

This API uses deep learning to generate face embedding 128 dimension vector using Keras on top of tensorflow.
Implementation is "FaceNet: A Unified Embedding for Face Recognition and Clustering". Inception-ResNet-v2 model.

Use 128 dimension vectors of different faces and match by calculating euclidean distace for each face with specified threshold.

Prerequisites

pip install -r requirements.txt

Run the server

gunicorn api:app

Endpoint: /face-embedding/url
Example: localhost:8000/face-embedding/url

Example

Request

/face-embedding/url

curl -i -H 'Content-Type: application/json' \
-d '{"url":"https://i.imgur.com/oRa0KpU.jpeg"}' \
http://localhost:8000/face-embedding/url

Response

HTTP/1.1 201 CREATED
Server: gunicorn/20.0.4
Date: Mon, 20 Apr 2020 18:55:52 GMT
Connection: close
Content-Type: application/json
Content-Length: 2581

{"face_embedding":{"url":"https://i.imgur.com/oRa0KpU.jpeg","vector":[1.0644716024398804,-0.7762305736541748,-1.5508257150650024,0.09768518060445786,3.234025001525879,0.7380242943763733,0.7755756378173828,0.7783681154251099,1.7994431257247925,1.3094197511672974,0.11752720177173615,-0.9972988367080688,-1.387189269065857,-1.0655053853988647,0.6015282869338989,-0.6466184854507446,-0.41857847571372986,-0.10470283031463623,0.4276144504547119,0.7561772465705872,1.6428494453430176,0.7238189578056335,-0.4318429231643677,0.4909619688987732,0.6246815919876099,-0.16436511278152466,0.643584668636322,-1.6872013807296753,-0.41700923442840576,-0.7216029763221741,1.135551929473877,-0.08613882958889008,-0.5983056426048279,-0.36481523513793945,-0.5517364740371704,0.3539840877056122,0.009662304073572159,-0.9530348181724548,-0.3983016014099121,-1.7624664306640625,-1.8313381671905518,0.34050893783569336,-0.5064315795898438,-0.6547110080718994,-0.1585574895143509,0.40287601947784424,-0.8792572617530823,0.8725411891937256,-0.5235766172409058,0.551956295967102,-1.616973876953125,0.7344838380813599,-2.1006178855895996,0.6636972427368164,0.3945094347000122,-1.370570182800293,0.07480968534946442,-0.19894933700561523,-0.5298669338226318,-0.7848787307739258,-0.6934372186660767,1.0431194305419922,1.1103864908218384,0.5065039396286011,0.2555731236934662,1.686985731124878,-0.7395192384719849,1.5575898885726929,-0.5048502683639526,0.39747563004493713,0.5613957643508911,-1.6452418565750122,-1.4124250411987305,-0.37187397480010986,0.2637156844139099,0.044945698231458664,-0.8186399936676025,-0.3490537405014038,-1.9461230039596558,2.091184377670288,-1.876999855041504,0.02519148588180542,0.08719244599342346,0.16580398380756378,0.3964640200138092,0.5649416446685791,-1.141135573387146,0.4675142168998718,1.0416184663772583,-1.6518527269363403,1.2884585857391357,0.11591645330190659,0.150620698928833,0.7876836061477661,0.09340327233076096,-0.17070399224758148,-1.5061938762664795,-0.24413225054740906,-0.4058179557323456,0.8020711541175842,-0.6087077856063843,-0.03180114924907684,-0.09178069233894348,-1.613738775253296,-0.155301034450531,-0.5883173942565918,0.4585244953632355,-0.6013964414596558,1.2576067447662354,0.4427909851074219,-0.9436922073364258,-1.8521229028701782,0.2701607942581177,-0.5950731635093689,0.5576406717300415,0.9674991369247437,-1.1031535863876343,0.8175444006919861,0.2308807671070099,0.8598411083221436,-0.2858640253543854,1.9551265239715576,0.2581416964530945,-0.654570460319519,0.7529764771461487,-1.4225571155548096,0.16272659599781036,0.09841877222061157]}}

License

This project is licensed under the MIT License - see the LICENSE.md file for details.