This project is an example application using Azure Functions to provide a REST interface for interacting with an Image Prediction Model from torchvision.
First you must create an azure account with the correct credentials.
And install Azure Core function tools
Install the project into the virtual environment as listed in the Contributing section below.
python3 -m venv .venv
source .venv/bin/activate (linux) or source .venv/Scripts/activate (windows)
pip install --no-cache-dir -r requirements.txt
Once in the venv you can run the application locally within the project folder:
func start --verbose
or within VS Studio:
Start the emulator by pressing F1 and selection Azurite: Start
Run locally by going to Run & Debug panel and pushing F5 and then sending a request
Test the app by opening a browser and running the following request: http://localhost:7071/api/predict?img=https://raw.githubusercontent.com/gvashishtha/functions-pytorch/master/resources/assets/Bernese-Mountain-Dog-Temperament-long.jpg
This project also provides a frontend splash page that you can run locally to interface with the application.
Open the file in frontend/index.html which will open a place for you to submit to an image URL for classification.
Try this one to see a bald eagle classified: https://i.natgeofe.com/k/afe51970-80c6-46c3-b047-4c407c72d874/bald-eagle-closeup_4x3.jpg
To deploy the project see the publish-to-azure project below.
The deploy_app bash file to setup the Azure application structure.
Note that names will need to be changed to make them unique.
To deploy the latest code, run the following within the virutal environement:
func azure functionapp publish azImagePredictor --build local
func azure functionapp publish azImagePredictor --build-native-deps
Use the remote version of the statement above if there is issue with the build
If you deploy to Azure you can replace the http://localhost:7071 in the statements above with the deployed application URL: https://azimagepredictor.azurewebsites.net
Similarly there is a file frontend/index_deployed.html which points to the deployed URL above.
- https://learn.microsoft.com/en-us/azure/azure-functions/machine-learning-pytorch?tabs=bash
- https://learn.microsoft.com/en-us/azure/azure-functions/create-first-function-vs-code-python?pivots=python-mode-decorators#publish-the-project-to-azure
- https://learn.microsoft.com/en-us/azure/azure-functions/functions-bindings-http-webhook-trigger?tabs=python-v1%2Cin-process%2Cnodejs-v4%2Cfunctionsv2&pivots=programming-language-python#example
- https://learn.microsoft.com/en-us/azure/azure-functions/functions-reference-python?tabs=asgi%2Capplication-level&pivots=python-mode-decorators#development-options
- To contribute to the repository use the following:
git clone <ENTER SSH HERE>
cd reponame
git lfs install
git checkout -b feature/feature_name
git add <files>
git commit -m "Add my feature"
git push origin feature/feature_name