Grepy

This prototype project provides Caffe inference functionality via REST api.

Build

Grepy is provided as Docker container. So you need to build Docker container at first.

$ cd docker/caffe_cpu
$ sudo docker build -t local/caffe_cpu -f docker/caffe_cpu/Dockerfile.caffe_cpu .
$ cd ../grepy_cpu
$ sudo docker build -t local/grepy_cpu -f Dockerfile.grepy_cpu .

This leads following micro-service. This managed under supervisord.

[Client]--[Nginx]--[uWSGI]--[server.py(Flask)]--[pretrained CaffeNet(Caffe)]

And run Grepy container. You can remove the container at stopping continer using --rm option.

$ sudo docker run --rm -i -t -p 80:80 local/grepy_cpu

Usage

You need to prepare test picture. I used Caltech 101.

$ wget -O - http://www.vision.caltech.edu/Image_Datasets/Caltech101/101_ObjectCategories.tar.gz | tar xfz -
$ curl --form "image=@101_ObjectCategories/airplanes/image_0001.jpg" http://(YOUR_HOST_IP_ADDRESS)/classify


      {
        "result": [
            {
               "name": "n02690373 airliner",
               "score": "0.952552318573"
            },
            {
               "name": "n04552348 warplane, military plane",
               "score": "0.0273641180247"
            },
            {
               "name": "n04008634 projectile, missile",
               "score": "0.00465240329504"
            }
         ]
       }