|
Download README.md from SimWorld-AI/code4scene-t2s-scenes: direct link, hf CLI and curl.
- Browser
- Download file 1.57 kB
-
https://huggingface.co/datasets/SimWorld-AI/code4scene-t2s-scenes/resolve/main/README.md
- Command line
-
hf download hf://datasets/SimWorld-AI/code4scene-t2s-scenes/README.md
-
curl -L -o README.md https://huggingface.co/datasets/SimWorld-AI/code4scene-t2s-scenes/resolve/main/README.md
1.57 kB
| pretty_name: Code4Scene Research Track 3D Scenes | |
| task_categories: | |
| - other | |
| tags: | |
| - code4scene | |
| - 3d | |
| - gltf | |
| - text-to-scene | |
| # Code4Scene Research Track 3D Scenes | |
| Browser-ready exports of saved Text-to-Scene outputs from the Code4Scene Research Track. | |
| The interactive viewer and benchmark results are at https://simworld-ai-code4scene.static.hf.space/gallery/explore.html. | |
| The initial release covers the ten highest-ranked model configurations from the October 19, 2026 leaderboard snapshot. | |
| Verified scenes are added as exports become available. `index.json` records the available case/model pairs, per-case | |
| scores, reference cameras, file sizes, SHA-256 hashes, and Standard/High variants. Missing exports are not benchmark failures. | |
| ## Files and viewing | |
| Files are meshopt-compressed glTF 2.0 GLBs with WebP textures, using the existing Code4Scene packed binary format. | |
| High: `<case-id>/<model-key>.bin`. Standard: `<case-id>/<model-key>-standard.bin`. | |
| The public website contains the loader and unpacker. Standard reduces texture resolution and may simplify dense geometry. | |
| The original Unreal Engine evaluator renders remain accessible alongside the interactive scenes; browser lighting and | |
| baked materials can differ from those renders. | |
| Each scene is downloaded individually when opened. Consumers can pin a repository revision and verify the hash | |
| recorded in `index.json`. Model names and scene scores refer to the benchmark configurations, not to model weights. | |
| Project: https://github.com/SimWorld-AI/Code4Scene | |
| Paper: https://arxiv.org/abs/2609.36777 | |