Welcome to the documentation for the MOMA-LRG dataset!

The Multi-Object, Multi-Actor Dataset with Language Refined Graphs (MOMA-LRG) is an novel benchmark designed to develop highly general and interpretable video understanding models.

The dataset is a research project from the Stanford Vision and Learning Lab.

Note

The dataset and its API are currently under active development. For the source code, please navigate to this GitHub repository.

• Getting Started
  • Installation
  • Retrieving the dataset

• Example Usage
  • Creating a MOMA Object
  • Few-shot experiments
  • Working with the data

API Reference

• MOMA
• Annotation Structure
• Taxonomy
• Lookup