ICRA-Competitions2026

Isaac GR00T IKEA training report

Training metrics and checkpoints for Isaac GR00T model trained on Unitree G1 for IKEA assembly tasks, exported from Weights & Biases run with 31,750 training steps.

Downloads52
Likes1

Why This Matters for Physical AI

This dataset demonstrates large-scale vision-language model training for robotic manipulation, specifically for furniture assembly tasks, advancing multi-modal learning approaches for embodied AI.

Technical Profile

Modalities
rgblanguage
Robot Embodiments
Unitree G1
Environment
simulation
Task Types
assembly
Annotation Types
language_instructions
Part of the Isaac GR00T family

Access

Need custom rgb data?

Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.

Request a Sample Pack

Related Datasets