StereoDataset is a synthetic multiview stereo dataset rendered in Unreal Engine. Each scene folder contains synchronized RGB and depth videos per camera, along with scene-level metadata files. The dataset was created by stereo-dataset-neurips2026 and was last updated on May 4, 2026.
Use Cases
- Training multiview stereo depth estimation models based on synchronized RGB and depth video pairs.
- Benchmarking 3D reconstruction algorithms based on synthetic scenes with known camera trajectories.
- Developing novel view synthesis techniques based on multiview video data.
- Evaluating computer vision models for scene understanding based on combined RGB, depth, and metadata.
Strengths
- Provides synchronized RGB and depth video streams for each camera view.
- Includes scene-level metadata such as camera trajectories and baseline information.
- Scenes are synthetically generated with controlled combinations of map, character mesh, animation, spawn location, and camera trajectory.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count and total dataset size are unknown, which may limit suitability assessment.
- Data is synthetic, which may limit applicability to real-world scenarios.
Provenance
- Source
- huggingface
- Collection Method
- Synthetic rendering in Unreal Engine.
- Freshness
- Last updated 2026-05-04 02:48:03