PointMixer: MLP-Mixer for Point Cloud Understanding
This work addresses point cloud understanding for 3D vision applications, representing an incremental improvement by adapting an existing image-based method to a new domain.
The paper tackles the challenge of applying MLP-Mixer to point clouds, which are sparse and irregular, by proposing PointMixer, a universal operator that replaces token-mixing MLPs with a softmax function to mix features within and between point sets, achieving competitive or superior performance in tasks like semantic segmentation and classification compared to transformer-based methods.
MLP-Mixer has newly appeared as a new challenger against the realm of CNNs and transformer. Despite its simplicity compared to transformer, the concept of channel-mixing MLPs and token-mixing MLPs achieves noticeable performance in visual recognition tasks. Unlike images, point clouds are inherently sparse, unordered and irregular, which limits the direct use of MLP-Mixer for point cloud understanding. In this paper, we propose PointMixer, a universal point set operator that facilitates information sharing among unstructured 3D points. By simply replacing token-mixing MLPs with a softmax function, PointMixer can "mix" features within/between point sets. By doing so, PointMixer can be broadly used in the network as inter-set mixing, intra-set mixing, and pyramid mixing. Extensive experiments show the competitive or superior performance of PointMixer in semantic segmentation, classification, and point reconstruction against transformer-based methods.