harpreet

a robot can recognize every object in a room and still not know how to open a single drawer knowing there's a cabinet is scene understanding. knowing which handle to grab, which direction to pull, and that it's a drawer not a door is affordance understanding almost no dataset has it in 3D SceneFun3D: 14,800+ functional elements across 710 laser-scanned indoor scenes. every handle, knob, button, and switch labeled with its affordance, motion axis, motion type, and a natural language task description each scene's 3D point cloud with its iPad video recordings in fiftyone. the affordance annotations live in 3D and project into the video frames check it out here huggingface.co/datasets/Voxe…

at highway speed, anything beyond 75 m is already your problem. aiMotive Multimodal Dataset has 176 scenes of LiDAR, camera, and radar with 360-degree coverage, shot by day, at night, and in rain. 26,583 annotated frames, 14 object classes, and about 24% of boxes beyond 75 m. i parsed it into FiftyOne so you can scrub the LiDAR point cloud next to camera images with projected boxes. huggingface.co/datasets/Voxe… #PhysicalAI #AutonomousDriving

does your aerial segmentation model hold up as the drone climbs? SkyScenes is a synthetic aerial dataset of 33,600 UAV images from the CARLA simulator, with 28 semantic classes, instance masks, and depth. this subset has 280 samples, all ClearNoon at 0° pitch from 15m, 35m, and 60m. i parsed it into FiftyOne so you can scrub the RGB, depth, and segmentation slices of each scene. huggingface.co/datasets/Voxe… #PhysicalAI #ComputerVision

pedestrian forecasters learn from bird's-eye or static-camera views, so they never see where the walker was looking. EgoTraj has 75 participants from 14 nationalities walking self-chosen routes across sidewalks, crosswalks, and busy streets in Meta Quest Pro headsets. it carries 1024 x 1024 RGB video at 30 fps, 6DoF head pose, per-frame 3D eye gaze vectors, and VLM-generated scene annotations. use it to train first-person trajectory forecasting that conditions on gaze. github.com/yehiahmad/EgoTraj #PhysicalAI #ComputerVision

the same eight-sensor rig rides on a car, a boat, and a quadruped. OctoSense records stereo RGB and event cameras, thermal, an Ouster OS1-64 LiDAR, IMU, and RTK GPS on one PPS-locked clock. the 371 car sequences alone run 59 hours, and the full release is 382 sequences. i parsed 8 episodes into FiftyOne so you can filter by platform, split, or sensor coverage and scrub every stream together. huggingface.co/datasets/Voxe… #PhysicalAI #Robotics

at every timestamp, a robotaxi has 9 cameras, 7 LiDARs, and 3 radars to reconcile. kitscenes-multimodal from KIT is Frankfurt urban driving at 10 Hz, labeled with Lanelet2 HD-map polylines, projected LiDAR depth, and the future ego path. i parsed 680 frames from 4 validation scenes into FiftyOne so you can view every camera next to the fused LiDAR and radar cloud. huggingface.co/datasets/Voxe… #PhysicalAI #AutonomousDriving

exporting a planetary rover's MCAP recordings takes ages, because the script writes to disk constantly to avoid running out of memory. the BASEPROD scripts export the Bardenas Semi-Desert Planetary Rover Dataset, plot weather station data and LIBS spectra, and apply force/torque, bogie offset, and gyro offset corrections. input data is preserved, and corrected files carry CORRECTED in their names. run them on the dataset download linked in the README. the scripts are MIT licensed. github.com/spaceuma/baseprod #PhysicalAI #Robotics

a highway work zone with a 40 km/h limit and constant traffic gets watched by three LiDARs that sense differently. HighwayScene pairs an Ouster OS0 (rotating time-of-flight), an Aeva Aeries II (FMCW, with radial velocity per point), and a Blickfeld QB2 (solid-state). 30 episodes of about 20 seconds, built for a cross-sensor background subtraction benchmark. i parsed it into FiftyOne so you can play all three point clouds on one timeline and filter by train, val, or test. huggingface.co/datasets/Voxe… #PhysicalAI #ComputerVision

a SLAM benchmark is only as good as the ground-truth trajectory behind it. yuto-mms-multimodal pairs a 32-beam Velodyne HDL-32E LiDAR with a Ladybug 5 panoramic camera built from six sensors (8000x4000), plus GPS/IMU at 125 Hz. The source has no object labels, only raw synchronized streams and a photogrammetric 6-DOF ground-truth camera trajectory. i parsed 2 of its 4 sequences into FiftyOne so you can scrub the panorama next to the colorized LiDAR cloud. huggingface.co/datasets/Voxe… #PhysicalAI #Robotics

the hand leaves the frame. the 3D motion still has to come back. DreamHand uses video diffusion as a deterministic geometry encoder for 3D hand motion in egocentric video under occlusion. improvement on ARCTIC and HOT3D is 30 to 46%. outputs are 3D keypoints and trajectories. arxiv.org/abs/2608.20308 #PhysicalAI #Robotics

if you have a lidar on a tilt mount and an odometry stack that assumes it isn't welcome, this account is for you not a benchmark account. no "10 tools you need"

backgrounds, clutter, and lighting change from episode to episode, and both arms still have to finish the job. RoboTwin Unified has 27,500 simulated episodes, 6,075,103 frames, and 50 dual-arm tasks, with 3 camera streams and 14-DoF state and action vectors. i parsed it into FiftyOne so you can filter episodes by task text, length, or camera stream. huggingface.co/datasets/Voxe… #PhysicalAI #Robotics

a hospital, a Nordic harbor, a marsh, a Japanese city, an abandoned factory, and a cyberpunk downtown. one robot dataset. TartanGround from CMU AirLab has a trajectory in each, with a front camera, segmentation, per-frame lidar, ego pose, and IMU. i parsed six trajectories into FiftyOne so you can scrub camera, segmentation, and lidar together. huggingface.co/datasets/Voxe… #PhysicalAI #Robotics

a small quadruped keeps moving in the dark on event cameras and near-infrared. M-SEVIQ pairs 720p stereo event cameras and active NIR illumination with RealSense RGB-D, joint encoders, and RTK-GPS on a Unitree Go2, for agile locomotion under darkness. arxiv.org/abs/2601.02777 #PhysicalAI #Robotics

a tracked robot drove 9,760 metres of closed loops through two active construction sites. revisiting the same ground makes these recordings useful for loop closure and traversability work. an RGB-D camera, a Livox lidar, two IMUs, GNSS, and wheel encoders, over 1 hour 45 minutes. i parsed it into FiftyOne so you can scrub camera, depth, and lidar together. huggingface.co/datasets/Voxe… #PhysicalAI #Robotics

same dataset, seven robot bodies. a Unitree G1, two GR1 setups, three Panda setups, and an R1 Pro. NVIDIA's GR00T-X-Embodiment-Sim has them all. i parsed 42 episodes into FiftyOne, six per body, with camera streams and joint telemetry on one timeline so you can compare how each body does its tasks. huggingface.co/datasets/Voxe… #PhysicalAI #Robotics

indoors, depth, LiDAR, and the body cameras still have to agree. the Spot Telluride set is Boston Dynamics Spot traversal from ROS bags: RealSense D455 RGB-D, a 64-channel Ouster OS1 LiDAR, body cameras, and LIO-SAM odometry. huggingface.co/datasets/lori… #PhysicalAI #Robotics

Apple tracked 68 body and finger joints in 3D while people did 111 tabletop tasks wearing a Vision Pro. EgoDex pairs 1080p egocentric video at 30 Hz with a pose for every joint, camera extrinsics, and a language description. the test set alone is 3,243 episodes. ICLR 2026. i parsed it into FiftyOne so you can watch the 3D skeleton move next to the video. huggingface.co/datasets/Voxe… #PhysicalAI #ComputerVision

a humanoid assembling IKEA furniture records 35 motors at once, each with position, velocity, acceleration, and torque. the 2026 Humanoid IKEA Assembly Challenge adds seven camera streams per episode, including infrared pairs from both wrists, plus the annotated subtasks. this is a 6-episode, 31-minute subset. i parsed it into FiftyOne so you can scrub the torque plots next to the wrist cameras. huggingface.co/datasets/Voxe… #PhysicalAI #Robotics