- app/htmldirectory: This app is mounted in app/main.py
- home.html: Defines the image upload form (POST method)
- style.css: Defines how the page looks
- upload.js: Defines how the image preview appears after uploading
- app/main.py: Defines the FastAPI app that renders the HTML, initializes the model, handles the input of the uploaded image into the model, and returns the model prediction.
- app/run.sh: Start the FastAPI app using Uvicorn
- models/model.py: Defines the DeepBIQ model (PyTorch CNN + sklearn SVR) and how the uploaded image is passed through the model to generate the final prediction
- services/event_handler.py: Load the CNN and SVR upon app launch
- deepbiq_deploy: Contains trained CNN and SVR weights, as well as MOS scores of the combined dataset in order for us to scale the model predictions to a percentile later (more easy to interpret)
- Dockerfile: Container with all dependencies installed; runs run.sh to start the FastAPI app
- deepbiq.sh: Pull the Docker image
- Build and push image with
t=phoebezhouhuixin/iqa_service:latest && sudo docker build -t $t . && sudo docker push $t - Run deepbiq.sh with
sudo sh deepbiq.sh- deepbiq.sh specifies a volume map between the source files in the local directory (./app) and the Docker container's directory (/app). Thus, any changes made to main.py locally will be reflected automatically in the Docker container.
- Go to http://0.0.0.0:8000/ (localhost port 8000)
# Commands to run web app
docker build -t iqa_service .
docker run -v /home/jiayi/Documents/GitHub/iqa_service/app:/app -p 8000:8000 iqa_service