Improving Multimodal Distillation for 3D Semantic Segmentation under Domain Shift

TL;DR

This paper explores how to effectively use vision foundation models for unsupervised domain adaptation in 3D lidar semantic segmentation, emphasizing architecture choices and training strategies to improve generalization across different lidar sensors.

Contribution

It provides a comprehensive study on leveraging vision foundation models for lidar domain adaptation, highlighting architecture importance, pretraining benefits, and effective training protocols.

Findings

Pretrained lidar backbones significantly improve domain generalization.

Keeping the backbone frozen and training only an MLP head yields best results.

The proposed method achieves state-of-the-art performance across multiple domain shift scenarios.

Abstract

Semantic segmentation networks trained under full supervision for one type of lidar fail to generalize to unseen lidars without intervention. To reduce the performance gap under domain shifts, a recent trend is to leverage vision foundation models (VFMs) providing robust features across domains. In this work, we conduct an exhaustive study to identify recipes for exploiting VFMs in unsupervised domain adaptation for semantic segmentation of lidar point clouds. Building upon unsupervised image-to-lidar knowledge distillation, our study reveals that: (1) the architecture of the lidar backbone is key to maximize the generalization performance on a target domain; (2) it is possible to pretrain a single backbone once and for all, and use it to address many domain shifts; (3) best results are obtained by keeping the pretrained backbone frozen and training an MLP head for semantic segmentation. The resulting pipeline achieves state-of-the-art results in four widely-recognized and challenging settings. The code will be available at: https://github.com/valeoai/muddos.