Face Recognition

(Part of the Face Recognition with Deep Detect series.)

An interface for finding images within a category matching a pre-configured model.

This has been developed with pictures of faces in mind and pre-configured with the vgg_face model (trained on LFW), but could be used for any model and associated categories.

Installation

See the below steps for getting up and running.

Run a Deep Detect instance

Start a DD instance for predicting our images:-

docker run -p 8080:8080 --name dd beniz/deepdetect_cpu

Run an Elastic Search instance

Start an ES instance which indexes our DD results:-

docker run -p 9200:9200 -p 9300:9300 elasticsearch
-Enetwork.bind_host=0.0.0.0 -Ehttp.cors.enabled=true
-Ehttp.cors.allow-origin="*"

Create Classification Service

Follow the steps on Face Recognition with Deep Detect up until the "Providing Inputs" section.

This will create a face classification service for DD.

Preparing Your Data

Next we need to download and load our images into ES via DD.

Execute these in order:-

Configure

Copy util/config.json.example to util/config.json and configure the example parameters, pointing to directories for the images etc.

Load

load.js is the tool for loading the images into ES via DD (base64 encodes each image for submission).

Serve

serve-file.py provides a web server for serving up the images for the front-end from disk.

Run front-end

Configure .env

REACT_APP_API_URL - Path to ES API (http://localhost:9500 passed through to http://localhost:9200/ by proxy in package.json for development).
REACT_APP_IMG_URL - Path to serve service (http://localhost:9000)
PORT - The port to serve the front-end on. (9500)

Build + Run Front-End

yarn
yarn start

The front-end should now be running on your configured port.

Launch http://localhost:9500/ and try some searches!

See further for other random information.

gaving/face-recognition