world-engine
WEB-Dataset
WorldEngine Bimanual Dataset for Post-training
A large-scale, language-annotated real-robot bimanual manipulation dataset for
post-training robotics foundation models. It spans 90 everyday manipulation tasks
collected with a bimanual YAM follower arm teleoperated by a GELLO leader,
recording joint state, action, and three synchronized camera streams at 60 Hz.
Shared lineage, different story. This dataset shares its hardware, teleoperation
setup, and recording pipeline with the… See the full description on the dataset page: https://huggingface.co/datasets/WorldEngineAI/WEB-Dataset.WorldEngine
The missing infrastructure for Physical AI post-training in AD. Open-source. Production-validated.
Joint effort by OpenDriveLab at The University of Hong Kong, Huawei Inc. and Shanghai Innovation Institute (SII).
Highlights
A post-training framework for Physical AI: Systematically addresses the long-tail safety-critical data scarcity problem in autonomous driving.
Data-driven long-tail discovery: Failure-prone scenarios are automatically identified from… See the full description on the dataset page: https://huggingface.co/datasets/OpenDriveLab/WorldEngine.worldengine-counterfactual-lead-pair-ladder-v9
WorldEngine counterfactual lead-distance pair ladders v9
This public dataset contains 65,131 rendered counterfactual cases
organized into 8,491 same-history action/future groups. Each
case has 4 historical and 8 future CAM_F0 frames plus WorldEngine metadata.
Archives preserve complete pair groups. The audit bundle contains group JSON
manifests, receipts, validation reports and the completion contract. Large
generator-intermediate subset PKLs are intentionally excluded because… See the full description on the dataset page: https://huggingface.co/datasets/WeiXiCZ/worldengine-counterfactual-lead-pair-ladder-v9.worldengine-counterfactual-accel-balanced-v8
WorldEngine counterfactual acceleration-balanced v8
This repository contains 12,900 validated front-camera counterfactual cases
used for CAST research. Each case contains 12 CAM_F0 frames (4 history + 8 future)
and its WorldEngine/NAVSIM-style metadata with ego trajectory information.
Cases: 12,900
Images: 154,800
Categories: {"adjacent_vehicle": 417, "part_minimum_fill": 10918, "salvage_out_of_grid": 1369, "stationary_vehicle": 162, "stopped_queue": 34}
Generator specification… See the full description on the dataset page: https://huggingface.co/datasets/WeiXiCZ/worldengine-counterfactual-accel-balanced-v8.worldengine-ingrid-8635-portable-v1
WorldEngine in-grid 8,635 portable raw samples
This public package contains exactly 8,635 selected in-grid samples.
Each sample includes a model-independent INTERACT PKL, the original 12-frame
WorldEngine metadata, and 12 CAM_F0 images (4 history + 8 future).
After downloading the complete repository, run:
bash extract_all.sh DESTINATION
The extracted interact_raw/*.pkl files use paths relative to themselves and
can be consumed by CAST's slot-action-cache-v4 builder.
world_engine_plate_v4This dataset was created using LeRobot.
Dataset Structure
meta/info.json:
{
"codebase_version": "v2.1",
"robot_type": "pipper",
"total_episodes": 200,
"total_frames": 101016,
"total_tasks": 1,
"total_videos": 600,
"total_chunks": 1,
"chunks_size": 1000,
"fps": 60,
"splits": {
"train": "0:200"
},
"data_path": "data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet",
"video_path":… See the full description on the dataset page: https://huggingface.co/datasets/changhaonan/world_engine_plate_v4.