Image Detection with Keras and Wave

Using open source Keras and Grad-CAM the user can upload their own photos and get explainable predictions of what a standard, pre-trained model thinks is in the image. This app makes these open source tools more accessible by adding a easy-to-use front end.

This app can be used as a template for Wave App developers on how to integrate third-party modeling frameworks into your own apps.

Local Development

Create a python environment in the home app directory and install requirements. Python 3.7 is recommended but not required.

python3 -m venv venv
./venv/bin/pip install -r requirements.txt

When running this application locally, ensure that you are using the parameter for pointing to the images/ folder so that these files will be automatically available in the Wave serve.

H2O_WAVE_PUBLIC_DIR="/images/@/Users/mtanco/h2o_wave/apps/wave-grad-cam/images" ./venv/bin/wave run src.app

Visit the app at http://localhost:10101

Publish to the AI App Store

After configuring your h2o CLI to a specific environment, you can upload this app using

h2o bundle import