Rotation invariance is a fundamental requirement of point cloud processing when input point clouds are not aligned. Many non-invariant networks performing well on aligned point clouds do not perform equivalent to rotated ones. Thus non-invariant and invariant networks are developed separately and only benefit a little from each other, leading to repetitive and wasteful research efforts. In this paper, we aim to bridge this gap and incorporate rotation invariance with non-invariant networks for point clouds. To this end, we propose a novel rotation invariant learning method based on efficient invariant poses (EIPs). EIPs do not rely on novel features or operations. Instead, they only rotate input point clouds into invariant poses and apply non-invariant networks in feature processing. As the name implies, EIPs have negligible complexities (efficient) and solid theoretical foundations (invariant). Experimental results demonstrate that EIPs have competitive performances on several tasks. Without using new features or operations, EIPs yield the best results on ScanObjectNN (PB T50 RS) classification and ShapeNetPart segmentation task. Our code is available at: https://github.com/JaronTHU/EIP.

Point clouds play an important role in 3D analysis, which has broad applications in robotics and autonomous driving. The pre-training fine-tuning paradigm has shown great potential in the point cloud domain. Full fine-tuning is generally effective but leads to a heavy storage and computational burden, which becomes inefficient and unacceptable as the size of pre-trained models scales. Although efficient fine-tuning approaches have significant progress in other domains, they generally perform worse for point clouds. To overcome this dilemma, we revisit the official Point-MAE implementation and find the critical role of aggregation in fine-tuning performances. Inspired by such discoveries, we propose a novel dynamic aggregation (DA) method to replace previous static aggregation like mean or max pooling for pre-trained point cloud Transformers. Besides standard metrics such as accuracy or mIoU, we evaluate the number of tunable parameters and additional FLOPs for a fair comparison of our method to different fine-tuning approaches. We construct several DA variants and validate them through extensive experiments. Experimental results demonstrate that DA has competitive performances against full fine-tuning and other efficient fine-tuning approaches. The code is publicly available at https://github.com/JaronTHU/DynamicAggregation.

Existing permutation-invariant methods can be divided into two categories according to the aggregation scope, i.e. global aggregation and local one. Although the global aggregation methods, e. g., PointNet and Deep Sets, get involved in simpler structures, their performance is poorer than the local aggregation ones like PointNet++ and Point Transformer. It remains an open problem whether there exists a global aggregation method with a simple structure, competitive performance, and even much fewer parameters. In this paper, we propose a novel global aggregation permutation-invariant network based on dual MLP dot-product, called DuMLP-Pin, which is capable of being employed to extract features for set inputs, including unordered or unstructured pixel, attribute, and point cloud data sets. We strictly prove that any permutation-invariant function implemented by DuMLP-Pin can be decomposed into two or more permutation-equivariant ones in a dot-product way as the cardinality of the given input set is greater than a threshold. We also show that the DuMLP-Pin can be viewed as Deep Sets with strong constraints under certain conditions. The performance of DuMLP-Pin is evaluated on several different tasks with diverse data sets. The experimental results demonstrate that our DuMLP-Pin achieves the best results on the two classification problems for pixel sets and attribute sets. On both the point cloud classification and the part segmentation, the accuracy of DuMLP-Pin is very close to the so-far best-performing local aggregation method with only a 1-2% difference, while the number of required parameters is significantly reduced by more than 85% in classification and 69% in segmentation, respectively. The code is publicly available on https://github.com/JaronTHU/DuMLP-Pin.

Place recognition technology can eliminate cumulative errors and thus plays a vital role in autonomous driving. In this paper, the composite feature of point cloud and image data is obtained by multi-modal feature fusion, thereby improving positioning accuracy. Semantic features, instance features, topological features, and image texture features are integrated to obtain comprehensive features, presenting strong robustness and complex scene expression abilities. Topological features consist of intra-class features and inter-instance features, which allow users to obtain more comprehensive scene structure information. The place recognition methods of data-level fusion and feature-level fusion based on point cloud and image are compared. This paper verifies the proposed method on SemanticKitti and nuScenes datasets. The results show that it outperforms state-of-the-art place recognition methods.