Classification Model for the Fashion-MNIST Dataset.

The API can be executed using Docker or not.

1. run docker compose

   docker compose up

   or use -d to keep the container running in background

   docker compose up -d

1. Install the requirements

   pip install -r backend/requirements.txt

2. Run the API:

   uvicorn backend.main:app

1. To use the API, you need to do a POST request at localhost:8000/model/prediction containing the image on file key at the body.

2. The response is a JSON file with the following keys: 2.1. class_id: Prediction resul class id (0 to 9); 2.2. class_name: Prediction result class name; 2.3. confidence: Confidence of model for the prediction (0 to 1); 2.4. probs: List of all model prediction probabilities; 2.5. labels: List of the classes name; 2.6. filename: Name of the file used on prediction.

• backend: API implementation;
• models: Model used on the API;
• train: Python Notebook used to train the models.

By Matheus de Andrade Silva