Either download data from Kaggle directly and extract it to this folder:
https://www.kaggle.com/datasets/gpiosenka/100-bird-species
or run 00_setup.ipynb
Run 01_quick_train.ipynb to fine-tune a ResNet34 model with the provided data.
Run 02_to_onnx.ipynb to convert the fine-tuned model to ONNX which can be deployed on RHOAI using OpenVINO.
Adapt the PREDICTION URL in streamlit/app.py and deploy the streamlit app for drag and drop scoring as shown in the screenshot.
ONNX conversion taken from: https://github.com/tkeyo/fastai-onnx See this nice blog post for details: https://dev.to/tkeyo/export-fastai-resnet-models-to-onnx-2gj7