:cityscape: ZAHA :cityscape:

To date the largest benchmark for facade semantic segmentation of point clouds

[Download] [Benchmark] [WACV25 Paper(pre-print)] [More]

Password:

zahahadid

overview

classes

:star2: Highlights

601 mln annotated points in ZAHA (/ˌzɑːhɑː/)
Introducing LoFG: Level of Facade Generalization enabling hierarchical understanding of facades
Various architectural styles
In local and global (i.e., UTM) coordinate reference system (if struggling with the height transformation - apply 45.66 m vertical offset)
File names point to the official CityGML LoD2 building models of Bavaria (e.g., DEBY_LOD2_4959462 -> gml_id=DEBY_LOD2_4959462) [visualisation]
The settings file for adding your own data

:mag_right: Facade Semantic Segmentation Results at LoFG3

LoFG stands for Level of Facade Generalization. See more in the ZAHA paper introducing the concept here.

	PointNet	PointNet++	Point Transformer	DGCNN
OA	59.9	66.4	75.0	71.1
P	46.1	37.8	52.7	53.6
R	42.2	35.9	54.7	45.8
F1	38.7	34.8	52.1	44.5
IoU	26.4	25.6	41.6	33.4
Class scores
wall	61.1	68.5	76.8	83.8
window	25.6	26.3	43.1	64.1
door	13.5	7.8	19.8	21.6
balcony	25.1	0.0	77.5	66.7
molding	22.5	43.4	58.0	57.5
deco	0.0	0.0	5.0	0.0
column	22.4	33.4	0.0	37.2
arch	19.2	25.4	50.2	2.6
stairs	16.0	0.0	7.5	5.6
ground surface	12.0	0.0	24.4	21.3
terrain	53.5	53.5	57.6	68.0
roof	18.7	6.8	66.3	57.4
blinds	4.6	2.3	18.5	20.0
interior	59.7	69.1	72.8	88.0
other	42.7	47.1	70.6	74.1

:mag_right: Facade Semantic Segmentation Results at LoFG2

LoFG stands for Level of Facade Generalization. See more in the ZAHA paper introducing the concept here.

	PointNet	PointNet++	Point Transformer	DGCNN
OA	71.9	75.5	78.2	82.6
P	69.6	73.0	75.8	80.0
R	68.1	73.0	76.6	81.8
F1	68.1	72.6	76.1	80.4
IoU	55.8	59.8	63.9	68.5
Class scores
floor	92.3	87.6	90.7	92.1
decoration	26.2	47.1	47.0	70.0
structural	60.9	65.5	67.0	85.2
opening	28.2	27.2	36.0	66.2
other el.	71.2	71.6	78.9	88.8

:mortar_board: Publication

Please find the official publication introducing "ZAHA" at the WACV '25 here:

The paper [WACV25 - pending] [arxiv preprint]

and consider citing it:

@article{wysockietalZAHA,
    author = {Wysocki, O. and Tan, Y. and Froech, T. and Xia, Y. and Wysocki, M. and Hoegner, L. and Cremers, D. and Holst Ch.},
    title = {ZAHA: Introducing the Level of Facade Generalization and the Large-Scale Point Cloud Facade Semantic Segmentation Benchmark Dataset},
    booktitle = {Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)},
    year = {2025},

@misc{wysocki2024zahaintroducinglevelfacade,
      title={ZAHA: Introducing the Level of Facade Generalization and the Large-Scale Point Cloud Facade Semantic Segmentation Benchmark Dataset}, 
      author={Olaf Wysocki and Yue Tan and Thomas Froech and Yan Xia and Magdalena Wysocki and Ludwig Hoegner and Daniel Cremers and Christoph Holst},
      year={2024},
      eprint={2411.04865},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2411.04865}, 
}

:handshake: Acknowledgments

This work wouldn't be possible without Zhu et al. and their excellent TUM-MLS-2016 dataset. Thank you!
We are indebted to Jiarui Zhang, Yue Tan, Chenkun Zhang, and Prabin Gyawali for their diligent work in the annotation process. Thank you Hitachi group for releasing the Semantic Segmentation Editor as a user-friendly open-source tool that we could easily adapt to our needs.

Feel free to check out other facade semantic segmentation datasets, like the one devoted to cultural heritage of Matrone et al. and the inspiring ArCH dataset! Go ahead, check it out, and test your algorithms there too!

OloOcki / zaha