Maximilian Jaritz

Happy to share that our work in cross-modal UDA (xMUDA) has been accepted to CVPR'20. Kudos @max_jaritz. Check out the preprint at: arxiv.org/abs/1911.12676 #CVPR2020 #xMUDA #3dsegmentation

Today we release the code for MVPNet: github.com/maxjaritz/mvpn… It reproduces our results for 3D semantic segmentation on the ScanNet benchmark and also contains an efficient PointNet++ implementation. Method: 2D-3D fusion of local RGB-D views and the global point cloud.

Check out our new work xMUDA! Cross-Modal Unsupervised Domain Adaptation for 3D Semantic Segmentation - UDA approach that exploits multi-modality - 2D-3D cross-modal learning between images and point clouds - Works for many interesting adaptation scenarios arxiv.org/abs/1911.12676