Importing an Image Dataset into H2O Hydrogen Torch
- Predictive AI
- H2O Hydrogen Torch
This lesson walks through importing a preprocessed image regression dataset into H2O Hydrogen Torch, using a set of over 6,000 photos labeled with the total value of coins shown in each image. It covers selecting the dataset file from S3, then the Configure Dataset page, where you set the dataset name, problem type, train, validation, and test dataframes, and the label and image columns. After a brief sanity check, it shows the Sample Data Visualization page, which displays random training images with their label overlaid so you can confirm the data imported correctly before setting up an experiment.
Key takeaways
- How a preprocessed dataset needs to match a specific format depending on the problem type you're solving
- What fields the Configure Dataset page asks for, including the label column and the image column
- How the Sample Data Visualization page lets you confirm labels look correct before training a model
- Why an image regression problem type fits a task like predicting the total value of coins in a photo
Start owning your AI