Overview: H2O Hydrogen Torch
- Predictive AI
- H2O Hydrogen Torch
Overview
This video walks through H2O Hydrogen Torch, H2O.ai's platform for deep learning on image, text, and audio data, using an image object detection use case to identify weed heads in agricultural images. It covers importing data from S3, Kaggle, or direct upload, configuring dataset settings by problem type, and creating experiments with adjustable settings by experience level, from novice to master.
Key takeaways
- How to import a dataset into Hydrogen Torch from S3, Kaggle, or direct upload, and configure problem-type-specific settings
- How to verify a dataset loaded correctly using sample visualizations and data statistics
- How to create an experiment, choose an experience level from novice to master, and use grid search to test setting combinations
- What backbone, fold, loss function, and validation metric settings control during experiment setup
- How to read experiment results, including the loss chart, metrics tab, confusion matrix, and validation prediction insights
- How to generate predictions on new data and download logs, predictions, or a Python/MOJO scoring pipeline
Start owning your AI