README.md

GoogleNet Classifier

This is a wrapper of the Machine Learning lib Caffe running its classifier demo with GoogleNet model pre-trained. No other dependencies than docker.

Demo

classifier.irony.svc.tutum.io

Dependencies

• Docker (Use docker-machine if you are running on OSX or Windows)

How to use

Just start the docker container and map the port 5000 to a public port:

docker run -it -p 5000:5000 irony/caffe-docker-classifier

Open the docker ip in a web browser

open http://192.168.99.100:5000

or use the api:

curl http://192.168.99.100:5000/classify_url?imageurl=http://lorempixel.com/400/200/animals/2/

or POST to /classify_upload

Output:

  {
    "result": [
      true,
      [
        [
          "gorilla",
          "0.42251"
        ],
        [
          "baboon",
          "0.24627"
        ],
        [
          "patas",
          "0.13308"
        ],
        [
          "spider monkey",
          "0.06061"
        ],
        [
          "macaque",
          "0.05365"
        ]
      ],
      [
        [
          "primate",
          "2.02654"
        ],
        [
          "anthropoid ape",
          "1.33458"
        ],
        [
          "ape",
          "1.30788"
        ],
        [
          "monkey",
          "1.27961"
        ],
        [
          "great ape",
          "1.22666"
        ]
      ],
      "4.565"
    ]
  }

Next steps

• Provide arguments for using different models
• Test GPU optimized environment
• Remove unneccessary dependencies

Development

Just clone this repo and use this command to link the local app.py to the container:

docker build -t caffe .
docker run -it -v $(pwd)/app.py:/opt/caffe/examples/web_demo/app.py caffe

Pull requests are welcome!

License

Please read the license from the pretrained GoogleNet model here, including source ImageNet rights: http://caffe.berkeleyvision.org/model_zoo.html