(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.
See the below steps for getting up and running.
Start a DD instance for predicting our images:-
docker run -p 8080:8080 --name dd beniz/deepdetect_cpu
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="*"
Follow the steps on Face Recognition with Deep Detect up until the "Providing Inputs" section.
This will create a face classification service for DD.
Next we need to download and load our images into ES via DD.
Execute these in order:-
Copy util/config.json.example
to util/config.json
and configure the
example parameters, pointing to directories for the images etc.
load.js
is the tool for loading the images into ES via DD (base64 encodes
each image for submission).
serve-file.py
provides a web server for serving up the images for the
front-end from disk.
REACT_APP_API_URL
- Path to ES API (http://localhost:9500 passed through to http://localhost:9200/ byproxy
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)
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.