Skip to the content.

BuildingNet v0.0 Benchmark Leaderboard

This is the official leaderboard of the BuildingNet v0.0 benchmark. Please email buildingnetwebmaster@gmail.com to add or update your results.

In your email please provide the following information in this format:

Method, Author list, Paper title, Conference, Link to paper
Untextured Point Cloud Track, Part IoU, Shape IoU, Class acc
Textured Point Cloud Track, Part IoU, Shape IoU, Class acc
Untextured Mesh Track, Part IoU, Shape IoU, Class acc
Textured Mesh Track, Part IoU, Shape IoU, Class acc

Untextured Point Cloud Track

Model Part IoU Shape IoU Class acc.
MinkowskiNet [2] 26.9 22.2 62.2




Textured Point Cloud Track

Model Part IoU Shape IoU Class acc.
MinkowskiNet [2] 29.9 24.3 65.5




Untextured Mesh Track

Model Part IoU Shape IoU Class acc.
MinkowskiNet2Triangle [2] 28.8 26.7 64.8
MinkowskiNet2Group [2] 33.1 36.0 69.9
BuildingNetGNN [1] 40.0 44.0 74.5




Textured Mesh Track

Model Part IoU Shape IoU Class acc.
MinkowskiNet2Triangle [2] 32.8 29.2 68.1
MinkowskiNet2Group [2] 37.0 39.1 73.2
BuildingNetGNN [1] 42.6 46.8 77.8



References

[1] Pratheba Selvaraju, Mohamed Nabail, Marios Loizou, Maria Maslioukova, Melinos Averkiou, Andreas Andreou, Siddhartha Chaudhuri, Evangelos Kalogerakis. BuildingNet: Learning to Label 3D Buildings. In Proc. ICCV, 2021
[2] Christopher Choy, JunYoung Gwak, and Silvio Savarese. 4D Spatio-Temporal ConvNets: Minkowski Convolutional Neural Networks. In Proc. CVPR, 2019