This project aims to classify Xmas trees. The dataset is composed of images of Xmas trees retrieved from the internet, reddit, facebook groups, ... With treediculous, you'll be able to get a real opinion on your Xmas tree.
The website is available at https://treediculous.fr
The resulting predictions are not meant to be taken seriously. The fact that defining
a Xmas tree as ugly or nice is purely subjective and depends on the annotator.
- ugly
- nice
- if the tree is too much decorated → ugly
- if the tree is too simple → ugly
- if the color of the tree is extravagant → ugly
- if the tree is ugly → ugly
- if the shape of the tree is strange → ugly
- if the tree is too small → ugly
label-studio
- API : FastAPI
- Frontend : React
- Deployment : Docker, Azure, Github Actions, Terraform
- MLOps : AzureML
Docker images are available at https://github.com/patacoing/treediculous/pkgs/container/treediculous
- API : treediculous:api-version
It contains the FastAPI app and the model to classify the trees.
- Frontend : treediculous:web-version
It contains the React app to display the predictions as well as Caddy.
- ML Pipeline environment : treediculous:pipeline
It contains the environment to run the ML pipeline in azure
- api : FastAPI app to infer the model
- webapp : React app to call the backend
- model : azureml pipeline, model training
Need to seperate containers in 2 container groups:
- Caddy
- API + Frontend because when we deploy a container, the entire group is recreated so caddy always tries getting a certificate