aidata Command Line Tool¶

aidata is the command line tool we use to download, transform, load, and split AI datasets across MBARI projects. It supports the same workflows described throughout this documentation and is especially useful when you need repeatable exports for training or bulk data operations.

Source code: mbari-org/aidata

Install¶

The quickest install path is:

pip install mbari-aidata

When to use aidata¶

Use aidata when you need to:

• Download data from Tator for training workflows from versioned datasets or combine versions using NMS
• Load SDCAT outputs or existing annotations into a project
• Convert data between formats such as COCO, CIFAR, PASCAL VOC, and YOLO
• Apply preprocessing or augmentation such as cropping, resizing or creating negative examples around your annotations
• Split datasets before training

aidata supports loading from VOC XML files and SDCAT CSVs, including bounding boxes, saliency scores, and class labels. See SDCAT CSV Format for the required columns and how to load a file.

If you only need a quick CSV export for analysis, we recommend the Tator web interface instead.

Commands¶

• aidata download - Download data to COCO, CIFAR, or PASCAL VOC format
• aidata load - Load SDCAT data, images, or video from a directory or text file
• aidata db - Manage your database
• aidata transform - Transform downloaded data and apply augmentations
• aidata split - Split your data before training

Run aidata -h to see the top-level help, or use a subcommand such as aidata download -h for command-specific options.

Updated: 2026-08-12