Learning Multi-View Aggregation In the Wild for Large-Scale 3D Semantic Segmentation
This work addresses the problem of efficient and accurate 3D semantic segmentation for applications in robotics and autonomous systems, representing a novel method rather than an incremental improvement.
The paper tackles the challenge of merging large-scale point clouds and images for 3D semantic segmentation by proposing an end-to-end trainable multi-view aggregation model that leverages viewing conditions to combine features from arbitrary images, achieving state-of-the-art results with 74.7 mIoU on S3DIS and 58.3 mIoU on KITTI-360.
Recent works on 3D semantic segmentation propose to exploit the synergy between images and point clouds by processing each modality with a dedicated network and projecting learned 2D features onto 3D points. Merging large-scale point clouds and images raises several challenges, such as constructing a mapping between points and pixels, and aggregating features between multiple views. Current methods require mesh reconstruction or specialized sensors to recover occlusions, and use heuristics to select and aggregate available images. In contrast, we propose an end-to-end trainable multi-view aggregation model leveraging the viewing conditions of 3D points to merge features from images taken at arbitrary positions. Our method can combine standard 2D and 3D networks and outperforms both 3D models operating on colorized point clouds and hybrid 2D/3D networks without requiring colorization, meshing, or true depth maps. We set a new state-of-the-art for large-scale indoor/outdoor semantic segmentation on S3DIS (74.7 mIoU 6-Fold) and on KITTI-360 (58.3 mIoU). Our full pipeline is accessible at https://github.com/drprojects/DeepViewAgg, and only requires raw 3D scans and a set of images and poses.