ONNX Model Inference with FastAPI

This is a small FastAPI application that has a /predict endpoint for classifiying images using the Resnet-18 ONNX model.

The model and the imagenet classes are downloaded when the server is started.

Run the Webserver locally

fastapi dev main.py

Test the API

curl -X POST "http://127.0.0.1:8000/predict" -F "file=@path/to/image.jpg"

Result:

{
  "predicted_class_index": 123
}