Three Pillars improving Vision Foundation Model Distillation for Lidar

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Puy, Gilles, Gidaris, Spyros, Boulch, Alexandre, Siméoni, Oriane, Sautier, Corentin, Pérez, Patrick, Bursuc, Andrei, Marlet, Renaud
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914685375217664
author Puy, Gilles
Gidaris, Spyros
Boulch, Alexandre
Siméoni, Oriane
Sautier, Corentin
Pérez, Patrick
Bursuc, Andrei
Marlet, Renaud
author_facet Puy, Gilles
Gidaris, Spyros
Boulch, Alexandre
Siméoni, Oriane
Sautier, Corentin
Pérez, Patrick
Bursuc, Andrei
Marlet, Renaud
contents Self-supervised image backbones can be used to address complex 2D tasks (e.g., semantic segmentation, object discovery) very efficiently and with little or no downstream supervision. Ideally, 3D backbones for lidar should be able to inherit these properties after distillation of these powerful 2D features. The most recent methods for image-to-lidar distillation on autonomous driving data show promising results, obtained thanks to distillation methods that keep improving. Yet, we still notice a large performance gap when measuring the quality of distilled and fully supervised features by linear probing. In this work, instead of focusing only on the distillation method, we study the effect of three pillars for distillation: the 3D backbone, the pretrained 2D backbones, and the pretraining dataset. In particular, thanks to our scalable distillation method named ScaLR, we show that scaling the 2D and 3D backbones and pretraining on diverse datasets leads to a substantial improvement of the feature quality. This allows us to significantly reduce the gap between the quality of distilled and fully-supervised 3D features, and to improve the robustness of the pretrained backbones to domain gaps and perturbations.
format Preprint
id arxiv_https___arxiv_org_abs_2310_17504
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Three Pillars improving Vision Foundation Model Distillation for Lidar
Puy, Gilles
Gidaris, Spyros
Boulch, Alexandre
Siméoni, Oriane
Sautier, Corentin
Pérez, Patrick
Bursuc, Andrei
Marlet, Renaud
Computer Vision and Pattern Recognition
Self-supervised image backbones can be used to address complex 2D tasks (e.g., semantic segmentation, object discovery) very efficiently and with little or no downstream supervision. Ideally, 3D backbones for lidar should be able to inherit these properties after distillation of these powerful 2D features. The most recent methods for image-to-lidar distillation on autonomous driving data show promising results, obtained thanks to distillation methods that keep improving. Yet, we still notice a large performance gap when measuring the quality of distilled and fully supervised features by linear probing. In this work, instead of focusing only on the distillation method, we study the effect of three pillars for distillation: the 3D backbone, the pretrained 2D backbones, and the pretraining dataset. In particular, thanks to our scalable distillation method named ScaLR, we show that scaling the 2D and 3D backbones and pretraining on diverse datasets leads to a substantial improvement of the feature quality. This allows us to significantly reduce the gap between the quality of distilled and fully-supervised 3D features, and to improve the robustness of the pretrained backbones to domain gaps and perturbations.
title Three Pillars improving Vision Foundation Model Distillation for Lidar
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2310.17504