Face Detection Web API is written on Go. The detection approach is powered by github.com/esimov/pigo library. The library gives good results out of the box and it's free of charge :)

The API spec is implemented in go-swagger 2.0.

Install go-swagger with brew (visit https://goswagger.io/install.html for more options):

$ brew tap go-swagger/go-swagger
$ brew install go-swagger

Get the app:

$ go get -u github.com/regeda/faced

API Spec

Run go generate to get "go" code if swagger.yml changed.

Run the web server:

$ ./run.sh

or

$ go run cmd/face-detection-app-server/main.go \
  --port=5555 \
  --pigo-cascade-file=internal/pigo/cascade/facefinder \
  --pigo-puploc-file=internal/pigo/cascade/puploc \
  --pigo-flploc-dir=internal/pigo/cascade/lps

Run go run cmd/face-detection-app-server/main.go --help for more options

Then navigate to http://127.0.0.1:5555/docs in your preferred browser.

Tests

Put your commands in generate.go file if you require mocks for tests. Then run go generate.

Face Detection Web API is a cloud-enabled web service. No 3d-party APIs required. You can pack the application into the Docker image and spin up containers in the Kubernetes cluster.

To reduce CPU utilization of the server, the web service can be run behind a caching reverse proxy (aka Squid, Varnish, Nginx). Because of the HTTP handler for faces detection is implemented as HTTP GET method and the detection algorithm is based on a static data model.