An HTTP service to accurately predict, given a photograph of a billboard, which ad (out of a database of registered images) is currently showing.

1. pip3 install -r requirements.txt
2. gunicorn wsgi:application

POST training image to http://<address>/register as an HTTP multipart argument named image. The service will return a unique ID associated with that image to be used later for identification.

Curl Example: curl -F image=@train/1.png http://127.0.0.1:8000/register

POST the image of the billboard to http://<address>/register, again as an HTTP multipart argument named image. The service will return the unique ID associated with registered ad that is most probably a match (the same ID that was returned when that ad was registered).

Curl Example: curl -F image=@test/1.png http://127.0.0.1:8000/predict

Run python3 test.py