Dataset Overview
Definition：GigaWorld-1 world model assessment data SetA large-scale benchmarking baseline data set for the generation of world models, designed to measure the four core categories of capabilities around dynamic landscape physical consistency, interactive object trajectory prediction, camera-view point transformation and long-term rollout stability, with JSON metadata and real values, which can be used directly to score and cross-versa comparisons of world model capabilities.
| Dataset Name | GigaWorld-1 world model assessment data Set |
| Total Data Volume | 50,000+clips Assessment Snippet |
| Visual Modality | RGB-D sequence + LiDAR Light Cloud, 1920 x 1080 @30FPS |
| EvaluationDimension | Physical Consistency / Interactive Trajection Forecast / Visual Point Transformation / Long-term Rollout Stability |
| True Marker | Object trajectory, exposure events, blocking relationships and landscape changes |
| Metadata | JSON Structured Description: scene, object, camera parameters, evaluation indicators |
| Scope of Application | Generating world model, video generation model, body imitators assessment |
Core Data Fields
clip_id- Assessment of the spectro-spectro-spectro-spectro-spectro-specific dimensionsrgb_d_seq— RGB-D frame sequence and time stamplidar- The aligned Lidar light cloudsgt_trajectory- Interactive object real-value trajectorycamera_params- Internal participation externals, changing of points of viewmetrics- Physical consistency/stability rating interface fields
Annotation and Quality Inspection
- Real value tracks are corrected by professional marking teams and reviewed by both key frames
- The calibration of indicators is aligned with the mainstream world model list, supporting the replication of results
- Automation of verification: full-scale check of the integrity of the clip, time stamp continuity and metadata consistency
- Provides assessment scripts and baseline results to facilitate horizontal comparison
Applicable Scenarios and Delivery
| Apply scene | Generating world model assessments, video-generated model physical reasonableness assessments, body imitation authentication, long-term prediction stability studies |
| Delivery Format | Video/dot cloud data package + JSON metadata + evaluate scripts and baseline results |
| Delivery Cycle | 4-6 Weeks |
| License Type | Commercial Use License + Data De-identification + Complete Authorization Documents |
Datasets in the Same Section
Need Samples or Customization？
The Changsha Langhui Information Technology Co., Ltd. DataAssetsAPI Platform supports flexible customization by scenario, embodiment form and scale。
Contact NowNeed a Customized Embodied AI Data Solution？
Supports flexible combination by scenario, embodiment form, sensor configuration and scale, and provides data sample preview and technical support。