Visualization & data tools

Research

LMFAO augments LeRobot demonstration videos with visual variations for robot learning.

Research

mekiki checks demonstration data for action-state mismatch, idle footage and camera synchronization issues.

Research

Mesh Twin provides a MuJoCo-based synthetic-data pipeline for mobile-robot navigation.

Research

Reinforcement learning experiments for the Microduck bipedal robot.

Open source

Mosaico - The data platform for Physical AI

Research

A way of gathering, aggregating, storing and visualizing statistics from arbitrary sources in a flexible and real-time safe way in ROS/ROS2

Research

Additional Helpers for the Oxford Radar RobotCar Dataset https://ori.ox.ac.uk/datasets/radar-robotcar-dataset

Research

RAIDEN is a toolkit for YAM robots that streamlines calibration, coordinated bimanual data collection, and dataset conversion.

Research

Visualization tooling for spatial data and robotics debugging.

Research

A ROS 2 pipeline for deploying a learned LEAP Hand grasping policy.

Open source

C++ robotics library for kinematics, dynamics, motion planning, and control.

Research

RoboData Doctor provides conformance checks and semantic comparisons for robot-learning datasets.

Research

Roboscribe-AF provides multi-agent annotation workflows for LeRobot demonstration datasets.

Research

Reading material on developmental robot learning and curricula.

Research

RoboTrace SDK logs robot episodes with synchronized video, sensor and action data for observability and evaluation.

Research

ROS 2 bag export tools for images, point clouds and other recorded data.

Open source

Framework to evaluate peformance of ROS 2

Open source

Tracing tools for ROS 2.

Open source

Recording and playback of robot topic data in ROS 2.

Open source

Browser-native visualization for embodied AI data (MCAP, ROS bag, HDF5)

Research

3D visualization tools for ROS 2 robot data and state.

Open source

Visualization of robotics and autonomy data encoded in XVIZ.

Research

This PyBullet project generates synthetic RGB-D grasp data and trains unseen-object manipulation baselines.

Research

ViKi converts human RGB-D demonstrations into retargeted robot motions and LeRobot datasets.