A MDS-SDC project
Our project contains three docker containers
- GradioApp for using a ML models to classify images
- Tensorflow Model Server providing the image classifier
- Covid Dashboard build with Mercury
The docker container can be used during development on a local docker desktop instance.
For CI/CD the docker container is 0) triggerd by a push of the new sources into this Github repository,
- build using Githab actions in this repo,
- pushed to DockerHub,
- deployed to Azure Services by
- pulling the images from DockerHub
for classifying images waits for an image as input and outputs the predictions together with the probabilities and below a generated wordcloud of the predictions in their relative importances.
Wordcloud in Python
https://pypi.org/project/wordcloud/\
Gradio as Docker image
https://github.com/njanakiev/minimal-gradio
TF-Serving using Huggingface models
https://huggingface.co/blog/tf-serving-vision
Run setup.py to generate model files for the docker image
docker-compose up -d
docker-compose stop