PSMix: Robust Point Cloud Recognition through Spectral Domain Mixing

While data augmentation is essential for robust point cloud recognition, conventional spatial mixup strategies often compromise geometric integrity by generating physically unrealistic samples. To overcome this limitation, we propose PSMix, which shifts the mixing paradigm to the spectral domain via the Spherical Harmonic Transform. Instead of simple coordinate interpolation, PSMix performs a rotation-aware hierarchical mixing on spectral coefficients. This approach explicitly preserves global structural properties while diversifying local details, achieving a balance that spatial methods struggle to maintain. Complementing this, we introduce an adversarial rotation optimization strategy to enforce invariance against challenging orientations. Extensive experiments on ModelNet-C and ScanObjectNN-C demonstrate that PSMix achieves state-of-the-art robustness, while also serving as an orthogonal plug-in that further boosts the performance of existing spatial strategies.

Recognizing 3D shapes accurately is important for things like robotics and virtual reality, but current techniques can struggle when the objects are rotated or slightly noisy. To improve this, we developed a new method called PSMix, which mixes 3D shape information in a way that keeps the overall structure intact while adding variety to the details. We also designed a strategy to make our method robust against tricky rotations. Tests on popular 3D shape datasets show that PSMix not only handles difficult scenarios better than previous methods, but can also enhance existing techniques when used together. This work helps computers understand 3D objects more reliably, even under challenging conditions.