Sparse Cross-scale Attention Network for Efficient LiDAR Panoptic Segmentation
This work addresses efficiency and accuracy issues in 3D LiDAR panoptic segmentation for autonomous driving applications, representing a strong domain-specific advancement.
The paper tackles the challenges of 3D LiDAR panoptic segmentation, specifically long-range dependency modeling for large objects and separation of close objects, by proposing SCAN, a sparse cross-scale attention network that achieves state-of-the-art performance on the SemanticKITTI dataset with real-time inference speed.
Two major challenges of 3D LiDAR Panoptic Segmentation (PS) are that point clouds of an object are surface-aggregated and thus hard to model the long-range dependency especially for large instances, and that objects are too close to separate each other. Recent literature addresses these problems by time-consuming grouping processes such as dual-clustering, mean-shift offsets, etc., or by bird-eye-view (BEV) dense centroid representation that downplays geometry. However, the long-range geometry relationship has not been sufficiently modeled by local feature learning from the above methods. To this end, we present SCAN, a novel sparse cross-scale attention network to first align multi-scale sparse features with global voxel-encoded attention to capture the long-range relationship of instance context, which can boost the regression accuracy of the over-segmented large objects. For the surface-aggregated points, SCAN adopts a novel sparse class-agnostic representation of instance centroids, which can not only maintain the sparsity of aligned features to solve the under-segmentation on small objects, but also reduce the computation amount of the network through sparse convolution. Our method outperforms previous methods by a large margin in the SemanticKITTI dataset for the challenging 3D PS task, achieving 1st place with a real-time inference speed.