Purge-Gate: Backpropagation-Free Test-Time Adaptation for Point Clouds Classification via Token Purging

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yazdanpanah, Moslem, Bahri, Ali, Noori, Mehrdad, Dastani, Sahar, Hakim, Gustavo Adolfo Vargas, Osowiechi, David, Ayed, Ismail Ben, Desrosiers, Christian
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908534825811968
author Yazdanpanah, Moslem
Bahri, Ali
Noori, Mehrdad
Dastani, Sahar
Hakim, Gustavo Adolfo Vargas
Osowiechi, David
Ayed, Ismail Ben
Desrosiers, Christian
author_facet Yazdanpanah, Moslem
Bahri, Ali
Noori, Mehrdad
Dastani, Sahar
Hakim, Gustavo Adolfo Vargas
Osowiechi, David
Ayed, Ismail Ben
Desrosiers, Christian
contents Test-time adaptation (TTA) is crucial for mitigating performance degradation caused by distribution shifts in 3D point cloud classification. In this work, we introduce Token Purging (PG), a novel backpropagation-free approach that removes tokens highly affected by domain shifts before they reach attention layers. Unlike existing TTA methods, PG operates at the token level, ensuring robust adaptation without iterative updates. We propose two variants: PG-SP, which leverages source statistics, and PG-SF, a fully source-free version relying on CLS-token-driven adaptation. Extensive evaluations on ModelNet40-C, ShapeNet-C, and ScanObjectNN-C demonstrate that PG-SP achieves an average of +10.3\% higher accuracy than state-of-the-art backpropagation-free methods, while PG-SF sets new benchmarks for source-free adaptation. Moreover, PG is 12.4 times faster and 5.5 times more memory efficient than our baseline, making it suitable for real-world deployment. Code is available at \hyperlink{https://github.com/MosyMosy/Purge-Gate}{https://github.com/MosyMosy/Purge-Gate}
format Preprint
id arxiv_https___arxiv_org_abs_2509_09785
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Purge-Gate: Backpropagation-Free Test-Time Adaptation for Point Clouds Classification via Token Purging
Yazdanpanah, Moslem
Bahri, Ali
Noori, Mehrdad
Dastani, Sahar
Hakim, Gustavo Adolfo Vargas
Osowiechi, David
Ayed, Ismail Ben
Desrosiers, Christian
Computer Vision and Pattern Recognition
Test-time adaptation (TTA) is crucial for mitigating performance degradation caused by distribution shifts in 3D point cloud classification. In this work, we introduce Token Purging (PG), a novel backpropagation-free approach that removes tokens highly affected by domain shifts before they reach attention layers. Unlike existing TTA methods, PG operates at the token level, ensuring robust adaptation without iterative updates. We propose two variants: PG-SP, which leverages source statistics, and PG-SF, a fully source-free version relying on CLS-token-driven adaptation. Extensive evaluations on ModelNet40-C, ShapeNet-C, and ScanObjectNN-C demonstrate that PG-SP achieves an average of +10.3\% higher accuracy than state-of-the-art backpropagation-free methods, while PG-SF sets new benchmarks for source-free adaptation. Moreover, PG is 12.4 times faster and 5.5 times more memory efficient than our baseline, making it suitable for real-world deployment. Code is available at \hyperlink{https://github.com/MosyMosy/Purge-Gate}{https://github.com/MosyMosy/Purge-Gate}
title Purge-Gate: Backpropagation-Free Test-Time Adaptation for Point Clouds Classification via Token Purging
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.09785