One of the best things about ImageMonkey is our tight integration with Machine learning frameworks like Tensorflow and Mask RCNN.
You want to train your own image classifier based on the ImageMonkey dataset?

It's as easy as that:

CPU version

    docker pull bbernhard/imagemonkey-train:latest
    docker run -it bbernhard/imagemonkey-train:latest

GPU version

    docker pull bbernhard/imagemonkey-train:latest-gpu
    docker run --runtime=nvidia -it bbernhard/imagemonkey-train:latest-gpu

This will download and run a docker image where we've already configured tensorflow and MaskRCNN to work seamlessly with ImageMonkey. After you've started the docker container, use the monkey script to interact with ImageMonkey dataset.

Here's a list of all available commands:

      [email protected]:/# monkey --help
      usage: PROG [-h] {train,list-labels} ...

      positional arguments:
        train              train your own model
        list-labels        list all labels that are available at ImageMonkey

        optional arguments:
        -h, --help           show this help message and exit


Let's assume you want to train your image classifier on all images that are labeled with dog or cat. Then simply run

      monkey train --labels="cat|dog" --type="image-classification"

lean back and go get yourself a coffee ;-)

The script automatically downloads all ImageMonkey images that are labeled with dog or cat and uses transfer learning to retrain an existing image classifier (inception v3) on these labels.

Umm, wait, I don't want to train an image classifier - I rather want to train a neural net that is able to do object segmentation. Nothing easier than that. Simply change the type

      monkey train --labels="cat|dog" --type="object-segmentation"
and MaskRCNN will be used instead.


In case you want to access ImageMonkey via Python, have a look at our Python library. But keep in mind, the library is still in an alpha stage and the API might change at any point!

Fork us on Github