360BEV: Panoramic Semantic Mapping for Indoor Bird’s-Eye View

[IEEE/CVF WACV 2024] Official Dataset Repository for the 360BEV benchmark.

• Project Page: https://jamycheung.github.io/360BEV.html
• GitHub Repository: https://github.com/InSAI-Lab/360BEV
• Paper: arXiv:2303.11910
• Model Checkpoints: InSAI-Lab/360BEV
• Collection: 360BEV & 360Mapper (WACV 2024)

---

Dataset Description

In this work, mapping from 360° panoramas to BEV semantics, the 360BEV task, is established for the first time to achieve holistic representations of indoor scenes in a top-down view. Instead of relying on narrow-FoV image sequences, a panoramic image with depth information is sufficient to generate a holistic BEV semantic map.

To benchmark 360BEV, we present two indoor datasets:

1. 360BEV-Stanford (Stanford-360): Built on Stanford2D3D, covering 6 indoor areas with egocentric panoramic images, 13 semantic classes, and allocentric top-down BEV semantic maps.
2. 360BEV-Matterport (Matterport-360): Built on Matterport3D, covering 86 buildings and 2,030 rooms with 20 semantic categories.
3. 360FV-Matterport: Front-view RGB, depth, and semantic segmentation panorama sequences for Matterport3D.

Dataset Statistics

Dataset	Split	Scene	Room	Frame	Category
360BEV-Stanford	Train	5	215	1,040	13
	Val	1	55	373	13
	Total	6	270	1,413	13
360BEV-Matterport	Train	61	--	7,829	20
	Val	7	--	772	20
	Test	18	--	2,014	20
	Total	86	2,030	10,615	20

Dataset Structure

After unpacking the archives into the data/ directory:

data/
├── Stanford2D3D
│   └── area_[1|2|3|4|5a|5b|6]
│       ├── rgb/*.png
│       └── semantic/*.png
│
├── 360BEV-Stanford
│   ├── bev/
│   │   └── *.h5 
│   ├── train/
│   │   └── *.png 
│   └── val/
│       └── *.png
│
├── 360BEV-Matterport
│   ├── bev/
│   │   └── *.h5 
│   ├── train/
│   │   └── *.png 
│   ├── val/
│   │   └── *.png
│   └── test/
│       └── *.png
│
└── 360FV-Matterport
    ├── 17DRP5sb8fy/
    │   ├── depth/*.png
    │   ├── rgb/*.png
    │   └── semantic/*.png   
    └── ...

Quick Start & Download

Using Hugging Face CLI (hf)

# Download entire dataset repository
hf download InSAI-Lab/360BEV --repo-type dataset --local-dir ./data

# Or download individual archives
hf download InSAI-Lab/360BEV 360BEV-Stanford.tar.gz --repo-type dataset --local-dir ./data
hf download InSAI-Lab/360BEV 360BEV-Matterport.tar.gz --repo-type dataset --local-dir ./data
hf download InSAI-Lab/360BEV 360FV-Matterport.tar.gz --repo-type dataset --local-dir ./data

Unpacking Archives

cd ./data
tar -xzf 360BEV-Stanford.tar.gz
tar -xzf 360BEV-Matterport.tar.gz
tar -xzf 360FV-Matterport.tar.gz

Citation

@inproceedings{teng2024_360bev,
  title={360BEV: Panoramic Semantic Mapping for Indoor Bird's-Eye View}, 
  author={Teng, Zhifeng and Zhang, Jiaming and Yang, Kailun and Peng, Kunyu and Shi, Hao and Reiß, Simon and Cao, Ke and Stiefelhagen, Rainer},
  booktitle={2024 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
  year={2024}
}

License

This dataset is released under the Apache-2.0 License.