datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
farm_uf850_home_fullfarm_uf850_home_v2farm_uf850_bottle
FARM UF850 — Bottle Pick-and-Place with a UFactory UF850
LeRobot-format dataset of teleoperated single-arm manipulation collected
on a UFactory UF850 (6-DoF arm with parallel gripper) using a Meta
Quest as the teleop controller. 200 episodes across 2 natural-language
tasks: moving a bottle between a desk and a box.
Collected with the FARM edge agent (repository)
as a CS153 final project. Intended for a full fine-tune of π0.5.
Quick stats
Episodes
200… See the full description on the dataset page: https://huggingface.co/datasets/NoahWeiss/farm_uf850_bottle.farm_uf850
FARM UF850 — Pick-and-Place with a UFactory UF850
LeRobot-format dataset of teleoperated single-arm manipulation collected
on a UFactory UF850 (6-DoF arm with parallel gripper) using a Meta
Quest as the teleop controller. 66 episodes spanning 17 distinct natural-
language tasks, primarily cube manipulation (pick the cube and place on
a box) including occlusion and failure-recovery scenarios.
Collected with the FARM edge agent (repository)
as a CS153 final project.… See the full description on the dataset page: https://huggingface.co/datasets/NoahWeiss/farm_uf850.farm_uf850_multiobject
farm_uf850_multiobject
424 teleoperated demonstrations of a UFactory UF850 arm doing 4 pick-and-place
tasks with different objects, in LeRobot
v2.0 format. Collected with the FARM VR-teleop harness (Meta Quest → ROS-TCP → π0.5).
Intended for π0.5 GSE-robust (domain-randomized) fine-tuning.
Summary
Episodes
424
Frames
129,067 @ 30 fps (~72 min)
Tasks
4
Robot
uf850 (6 joints + parallel gripper)
Cameras
base, wrist (640×480, h264)
Tasks… See the full description on the dataset page: https://huggingface.co/datasets/NoahWeiss/farm_uf850_multiobject.uf850_cube_cup_cleanuf850-vr-teleop-hang-the-cup-rlds-noop-filtered
UF850 VR Teleoperation Hang The Cup RLDS 10 Hz No-Op Filtered
Public OpenVLA-OFT/VLA-Adapter-compatible RLDS/TFDS trajectories derived from
CASE-Lab/uf850-vr-teleop-hang-the-cup-raw.
The raw release is endpoint-resampled to a local, checksum-pinned 10 Hz
intermediate and then filtered for idle actions. The final published release
keeps the original 10 Hz endpoint action semantics.
Task
Hang the yellow cup on the cup rack.
Instructions stored in… See the full description on the dataset page: https://huggingface.co/datasets/CASE-Lab/uf850-vr-teleop-hang-the-cup-rlds-noop-filtered.uf850-vr-teleop-hang-the-cup-raw
Hang The Cup — UF850 Quest VR Raw Data
Raw UF850 teleoperation trajectories collected with a Quest VR controller and
two Intel RealSense cameras. Published trajectories are not resampled, cropped,
action-normalized, or converted to RLDS/LeRobot. Camera frames were captured as
YUYV and stored as JPEG at collection time.
Task and prompt
Every episode uses the same English task instruction:
hang the yellow cup on the cup rack
The prompt is an intended collection… See the full description on the dataset page: https://huggingface.co/datasets/CASE-Lab/uf850-vr-teleop-hang-the-cup-raw.uf850-vr-teleop-press-buttons-rlds-noop-filtered
UF850 Quest VR Press Buttons — RLDS No-Op Filtered
OpenVLA-OFT-compatible RLDS/TFDS trajectories derived from the validated
UF850 Press Buttons raw dataset.
Idle steps were removed before image processing; retained actions keep their
original adjacent-sample meaning and are not recomputed after filtering.
Task instruction
Every step uses the same English instruction:
First press the blue button, then press the red one, and finally press the orange one… See the full description on the dataset page: https://huggingface.co/datasets/AAyano/uf850-vr-teleop-press-buttons-rlds-noop-filtered.uf850-vr-teleop-place-corn-in-bowl-rlds-noop-filtered
UF850 Quest VR Place Corn in Bowl — RLDS No-Op Filtered
OpenVLA-OFT-compatible RLDS/TFDS trajectories derived from the validated
UF850 Place Corn in Bowl raw dataset.
Idle steps were removed before image processing; retained actions keep their
original adjacent-sample meaning and are not recomputed after filtering.
Task instruction
Every step uses the same English instruction:
Place the corn in the orange bowl.
Summary
Item
Value… See the full description on the dataset page: https://huggingface.co/datasets/AAyano/uf850-vr-teleop-place-corn-in-bowl-rlds-noop-filtered.uf850-vr-teleop-place-corn-in-bowl-raw
UF850 Quest VR Place Corn in Bowl — Raw Data
Raw, untrimmed UF850 teleoperation trajectories collected with a Quest VR
controller and two Intel RealSense cameras. No trajectory trimming, resampling,
or action normalization has been applied. Camera frames were captured as YUYV
and stored as JPEG at collection time.
Task and prompt
Every episode uses the same English task instruction:
Place the corn in the orange bowl.
The canonical prompt is stored in… See the full description on the dataset page: https://huggingface.co/datasets/AAyano/uf850-vr-teleop-place-corn-in-bowl-raw.uf850-vr-teleop-press-buttons-raw
UF850 Quest VR Press Buttons — Raw Data
Raw, untrimmed UF850 teleoperation trajectories collected with a Quest VR
controller and two Intel RealSense cameras. No retained trajectory was trimmed,
resampled, or action-normalized. Camera frames were captured as YUYV and stored
as JPEG at collection time.
Task and prompt
Every episode uses the same English task instruction:
First press the blue button, then press the red one, and finally press the orange one
The… See the full description on the dataset page: https://huggingface.co/datasets/AAyano/uf850-vr-teleop-press-buttons-raw.