--- license: cc-by-nc-4.0 --- # Ctrl-Room dataset introduced in "Ctrl-Room: Controllable Text-to-3D Room Meshes Generation with Layout Constraints" [paper](https://arxiv.org/abs/2310.03602) | [project page](https://fangchuan.github.io/ctrl-room.github.io/) | [code](https://github.com/fangchuan/Ctrl-Room) This repository hosts the Ctrl-Room dataset, a collection of semantically annotated 3D scenes derived from the [Structured3D dataset](https://structured3d-dataset.org/). It contains 11,986 filtered rooms, including 4,961 bedrooms, 3,039 living rooms, 1,848 kitchens, 1,500 bathrooms, and 638 studies. Additionally, we also provide the scene descriptions following the format in [SceneFormer](https://arxiv.org/abs/2012.09793) for each room, which can be used for training text-to-3D generative models. ## Dataset Structure Outlines of the dataset files: ``` Ctrl-Room-Dataset/ ├── train/ # train split │ ├── bedroom/ # bedroom scenes | | ├── bbox_3d/ # 3D bounding box annotations (.json files) | | ├── cam_pos/ # camera positions | | ├── depth/ # panoramic depth maps | | ├── img/ # panoramic RGB images | | ├── label_cor/ # 2D room_layout annotations for each panoramic RGB image | | ├── quad_walls/ # 3D room_layout annotations (.json files) | | ├── room_type/ # room types | | ├── sem_layout_img/ # semantic layout panorama images | | ├── text_desc/ # text prompt for each room | | ├── text_desc_emb/ # pre-computed text embeddings using CLIP │ | └── train_dataset_stats.json # dataset statistic information │ ├── living_room/ # living room scenes │ ├── kitchen/ # kitchen scenes │ ├── study/ # study scenes │ └── bathroom/ # bathroom scenes ├── test/ # test split │ ├── bedroom/ # bedroom scenes │ ├── living_room/ # living room scenes │ ├── kitchen/ # kitchen scenes │ ├── study/ # study scenes │ └── bathroom/ # bathroom scenes ├── bedroom_splits.json # split file for bedroom ├── livingroom_splits.json # split file for living room ├── kitchen_splits.json # split file for kitchen └── study_splits.json # split file for study ``` ## Citation If you can make use of it in your own research, please be so kind to cite us. ``` @article{Ctrl-Room, title = {Ctrl-room: Controllable text-to-3d room meshes generation with layout constraints}, author = {Chuan Fang, Yuan Dong, Kunming Luo, Xiaotao Hu, Rakesh Shrestha, Ping Tan}, journal = {2025 International Conference on 3D Vision (3DV).}, year = {2025}, eprint = {2025: 692-701.}, archivePrefix = {IEEE}, primaryClass = {cs.CV} } ```