k8s-image-recognition

This repository was made for my Final engineering study project at Unilasalle Amiens France, thank you to the staff of the school as well as to the teachers for these 3 years which complete my 6 years of studies in the field of software and system/network engineering.

It concists in a simple automated k8s cluster build on AWS deploying an actix_web reverse proxy API to call an inference of google/pix2struct model that i trained for french road signs, stored in Huggingface.co.

CI\CD are building docker images resulting from the main branch's content.

How to Use :

kompose convert -f .\kube-docker-compose.yml
kubectl --kubeconfig=<Path to your configuration> apply -f <desirated-deployments>
...
kubectl expose deployment descrivizio001-web --type="NodePort" --port 80 --name=desrivizio001-web
kubectl expose deployment descrivizio001-api --type="NodePort" --port 8085 --name=descrivizio001-api
kubectl get pod
kubectl get service
kubectl get deployment

Linux/MacOS Image build requierments

echo "HUGGING_FACE_TOKEN=YOUR_HUGGINGFACE_TOKEN_TO_YOUR_MODEL" > ./descrivizio001/.env

Windows build requierments

New-Item -ItemType File -Path ".\descrivizio001\.env" -Value "HUGGING_FACE_TOKEN=YOUR_HUGGINGFACE_TOKEN_TO_YOUR_MODEL"