DAVIAN-Robotics/PointmapVLA-Pretrained

Robotics • Updated 3 days ago • 6 • 1

Note Pretrained on 76.3M pointmap robot frames + 10.7M 3D spatial VQA. Start here to fine-tune on your own robot.

DAVIAN-Robotics/PointmapVLA-RoboCasa

Robotics • Updated 3 days ago • 1

Note Fine-tuned on RoboCasa kitchen-24 (MG-300). 78.2% average success over 24 tasks.

DAVIAN-Robotics/PointmapVLA-LIBERO

Robotics • Updated 3 days ago • 2

Note Fine-tuned on the 4 LIBERO suites. 99.0% on LIBERO, 92.4% zero-shot on LIBERO-Plus.

Note RoboCasa MG-300 fine-tuning data: RGB (3 cameras), state, action. 7,200 episodes, LeRobot v3.0.

Note Depth videos + per-frame camera parameters for robocasa-MG_300-r6; pointmaps are built on the fly.

Note LIBERO fine-tuning data (4 suites): RGB (2 cameras), state, action. 1,693 episodes, LeRobot v3.0.

Note Depth videos + camera parameters for libero-r6; pointmaps are built on the fly.