https://github.com/joshbrew/xgboost_onnx_training_conversion

Image and CSV classification samples using ONNX with (bad) sample data.
https://github.com/joshbrew/xgboost_onnx_training_conversion

Last synced: 7 months ago
JSON representation

Image and CSV classification samples using ONNX with (bad) sample data.

• Host: GitHub
• URL: https://github.com/joshbrew/xgboost_onnx_training_conversion
• Owner: joshbrew
• License: lgpl-3.0
• Created: 2024-01-31T19:44:56.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2024-03-22T00:25:35.000Z (about 2 years ago)
• Last Synced: 2025-06-06T02:05:19.589Z (about 1 year ago)
• Language: Jupyter Notebook
• Size: 54.8 MB
• Stars: 1
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# XGBoost_ONNX_Notebooks

Image and CSV classification samples using ONNX with sample data. This demonstrates the powerful, minimal [XGBoost](https://xgboost.readthedocs.io/en/stable/) algorithm meant for portability while retaining competitive classification accuracy.

These are beginner-friendly python notebooks with plenty of instructions for creating ONNX models that can be loaded into your own software applications to get the power of offline machine learning into your phone or browser. You can see our results in action, running in milliseconds or even nanoseconds on WebGPU or WebGL at: [https://github.com/joshbrew/cameraId-wonnx-wasm](https://github.com/joshbrew/cameraId-wonnx-wasm)

Follow the instructions in the XGBoost subfolders for training the respective models on your local machine. 

They are fairly identical processes. Also find google colab links in those READMEs for trying it out online.

We're working on getting better datasets, they are homemade for fish identification from digital spectral data and images of filets. This is part of a project for creating image recognition tools for conservation purposes with funding from Schmidt Marine and the Oregon Coast Aquarium.

More info forthcoming.

## Credits

- .ipynb notebooks by Mateusz Stankiewicz  ([flyps.io](https://flyps.io)) and Joshua Brewster

- datasets by Joshua Brewster (WIP)