Efficient Point Transformer with Dynamic Token Aggregating for LiDAR Point Cloud Processing

TL;DR

This paper introduces DTA-Former, an efficient 3D Transformer model for LiDAR point cloud processing that reduces computational costs through dynamic token aggregation and adaptive sparsification, achieving high classification accuracy.

Contribution

The paper presents a novel DTA-Former architecture with learnable token sparsification, dynamic token aggregation, and iterative token reconstruction for improved efficiency and performance in 3D point cloud analysis.

Findings

Achieves state-of-the-art classification accuracy on LiDAR datasets.

Reduces computational complexity compared to existing 3D Transformer methods.

Demonstrates effective dense prediction with the proposed ITR and W-net architecture.

Abstract

Recently, LiDAR point cloud processing and analysis have made great progress due to the development of 3D Transformers. However, existing 3D Transformer methods usually are computationally expensive and inefficient due to their huge and redundant attention maps. They also tend to be slow due to requiring time-consuming point cloud sampling and grouping processes. To address these issues, we propose an efficient point TransFormer with Dynamic Token Aggregating (DTA-Former) for point cloud representation and processing. Firstly, we propose an efficient Learnable Token Sparsification (LTS) block, which considers both local and global semantic information for the adaptive selection of key tokens. Secondly, to achieve the feature aggregation for sparsified tokens, we present the first Dynamic Token Aggregating (DTA) block in the 3D Transformer paradigm, providing our model with strong aggregated features while preventing information loss. After that, a dual-attention Transformer-based Global Feature Enhancement (GFE) block is used to improve the representation capability of the model. Equipped with LTS, DTA, and GFE blocks, DTA-Former achieves excellent classification results via hierarchical feature learning. Lastly, a novel Iterative Token Reconstruction (ITR) block is introduced for dense prediction whereby the semantic features of tokens and their semantic relationships are gradually optimized during iterative reconstruction. Based on ITR, we propose a new W-net architecture, which is more suitable for Transformer-based feature learning than the common U-net design.