In the realm of 3D point clouds, the availability of large datasets is a challenge, which exacerbates the issue of training Transformers for 3D tasks.
In this work, we revisit the classical PointNet++ through a systematic study of model training and scaling strategies, and offer two major contributions.
Ranked #1 on Semantic Segmentation on S3DIS
TNAS performs a modified bi-level Breadth-First Search in the proposed trees to discover a high-performance architecture.
We then introduce a new Anisotropic Reduction function into our Separable SA module and propose an Anisotropic Separable SA (ASSA) module that substantially increases the network's accuracy.
Ranked #10 on Semantic Segmentation on S3DIS Area5
We combine Inception DenseGCN with NodeShuffle into a new point upsampling pipeline called PU-GCN.
Architecture design has become a crucial component of successful deep learning.
Ranked #3 on Node Classification on PPI
This work transfers concepts such as residual/dense connections and dilated convolutions from CNNs to GCNs in order to successfully train very deep GCNs.
Ranked #4 on 3D Semantic Segmentation on PartNet
Such a mixture problem is usually solved by a sequential solution (applying each method independently in a fixed order: DM $\to$ DN $\to$ SR), or is simply tackled by an end-to-end network without enough analysis into interactions among tasks, resulting in an undesired performance drop in the final image quality.