Smaller3d: Smaller Models for 3D Semantic Segmentation Using Minkowski Engine and Knowledge Distillation Methods

TL;DR

This paper introduces smaller 3D semantic segmentation models using Minkowski Engine and knowledge distillation, achieving significant size reduction with minimal performance loss on ScanNet V2.

Contribution

It applies knowledge distillation to sparse 3D models, enabling substantial size reduction while maintaining competitive accuracy.

Findings

Achieved 2.6% mIoU difference with 4x smaller models.

Achieved 8% mIoU difference with 16x smaller models.

Validated on ScanNet V2 dataset.

Abstract

There are various optimization techniques in the realm of 3D, including point cloud-based approaches that use mesh, texture, and voxels which optimize how you store, and how do calculate in 3D. These techniques employ methods such as feed-forward networks, 3D convolutions, graph neural networks, transformers, and sparse tensors. However, the field of 3D is one of the most computationally expensive fields, and these methods have yet to achieve their full potential due to their large capacity, complexity, and computation limits. This paper proposes the application of knowledge distillation techniques, especially for sparse tensors in 3D deep learning, to reduce model sizes while maintaining performance. We analyze and purpose different loss functions, including standard methods and combinations of various losses, to simulate the performance of state-of-the-art models of different Sparse Convolutional NNs. Our experiments are done on the standard ScanNet V2 dataset, and we achieved around 2.6\% mIoU difference with a 4 times smaller model and around 8\% with a 16 times smaller model on the latest state-of-the-art spacio-temporal convents based models.