SALA: Soft Assignment Local Aggregation for Parameter Efficient 3D Semantic Segmentation

TL;DR

This paper introduces SALA, a learnable neighbor-to-grid assignment method for 3D point cloud segmentation that achieves state-of-the-art results with significantly fewer parameters.

Contribution

It proposes a novel learnable soft assignment function for local aggregation, enabling parameter-efficient models for 3D semantic segmentation.

Findings

Achieves SOTA on S3DIS with 10x fewer parameters.

Demonstrates competitive results on ScanNet and PartNet.

Introduces flexible, layer-specific assignment functions.

Abstract

In this work, we focus on designing a point local aggregation function that yields parameter efficient networks for 3D point cloud semantic segmentation. We explore the idea of using learnable neighbor-to-grid soft assignment in grid-based aggregation functions. Previous methods in literature operate on a predefined geometric grid such as local volume partitions or irregular kernel points. A more general alternative is to allow the network to learn an assignment function that best suits the end task. Since it is learnable, this mapping is allowed to be different per layer instead of being applied uniformly throughout the depth of the network. By endowing the network with the flexibility to learn its own neighbor-to-grid assignment, we arrive at parameter efficient models that achieve state-of-the-art (SOTA) performance on S3DIS with at least 10$\times$ less parameters than the current reigning method. We also demonstrate competitive performance on ScanNet and PartNet compared with much larger SOTA models.