MVMO: A Multi-Object Dataset for Wide Baseline Multi-View Semantic Segmentation


Aitor Alvarez-Gila1,2
Joost van de Weijer2
Yaxing Wang2
Estibaliz Garrote1

1Tecnalia - Basque Research and Technology Alliance (BRTA)
2Computer Vision Center - Universitat Autònoma de Barcelona


(Under Review)

[Paper]
[Data (coming soon!)]
[Code (coming soon!)]


Abstract

We present MVMO (Multi-View, Multi-Object dataset): a synthetic dataset of 116,000 scenes containing randomly placed objects of 10 distinct classes and captured from 25 camera locations in the upper hemisphere. MVMO comprises photorealistic, path-traced image renders, together with semantic segmentation ground truth for every view. Unlike existing multi-view datasets, MVMO features wide baselines between cameras and high density of objects, which lead to large disparities, heavy occlusions and view-dependent object appearance. Single view semantic segmentation is hindered by self and inter-object occlusions that could benefit from additional viewpoints. Therefore, we expect that MVMO will propel research in multi-view semantic segmentation and cross-view semantic transfer. We also provide baselines that show that new research is needed in such fields to exploit the complementary information of multi-view setups.



Comparison of datasets suitable for multi/cross-view semantic segmentation (updated 2022-03-01)



Sample scenes


Paper

Aitor Alvarez-Gila, Joost van de Weijer, Yaxing Wang, Estibaliz Garrote.

MVMO: A Multi-Object Dataset for Wide Baseline Multi-View Semantic Segmentation

(Under review).

[bibtex]

Contact: Aitor Alvarez-Gila