Skip to content

Dataset · IPEC-COMMUNITY

Multimodal dataset for robotics and embodied AI

IPEC-COMMUNITY/EO-Data1.5M

Large-scale set of 1 to 10 million examples alternating images, text, and actions for training embodied AI.

What's inside

This is a large multimodal corpus (ranging from 1M to 10M records) alternating visual data, English text descriptions, and actions. The data is structured with an emphasis on temporal dynamics and causal relationships between what the agent sees, understands, and does. The dataset is designed to train models in complex reasoning within a physical environment.

What you can do with it

  • Train Vision-Language-Action (VLA) models for robot control
  • Train multimodal systems on visual question answering and video sequence analysis tasks
Tasks
Robotics, Visual question answering, Video text → text, Image text → text
Languages
English
Size
1M–10M rows
Downloads this month
20.3K
Likes
24

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("IPEC-COMMUNITY/EO-Data1.5M")