Learning Robust Anymodal Segmentor with Unimodal and Cross-modal Distillation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zheng, Xu, Xue, Haiwei, Chen, Jialei, Yan, Yibo, Jiang, Lutao, Lyu, Yuanhuiyi, Yang, Kailun, Zhang, Linfeng, Hu, Xuming
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909611580194816
author Zheng, Xu
Xue, Haiwei
Chen, Jialei
Yan, Yibo
Jiang, Lutao
Lyu, Yuanhuiyi
Yang, Kailun
Zhang, Linfeng
Hu, Xuming
author_facet Zheng, Xu
Xue, Haiwei
Chen, Jialei
Yan, Yibo
Jiang, Lutao
Lyu, Yuanhuiyi
Yang, Kailun
Zhang, Linfeng
Hu, Xuming
contents Simultaneously using multimodal inputs from multiple sensors to train segmentors is intuitively advantageous but practically challenging. A key challenge is unimodal bias, where multimodal segmentors over rely on certain modalities, causing performance drops when others are missing, common in real world applications. To this end, we develop the first framework for learning robust segmentor that can handle any combinations of visual modalities. Specifically, we first introduce a parallel multimodal learning strategy for learning a strong teacher. The cross-modal and unimodal distillation is then achieved in the multi scale representation space by transferring the feature level knowledge from multimodal to anymodal segmentors, aiming at addressing the unimodal bias and avoiding over-reliance on specific modalities. Moreover, a prediction level modality agnostic semantic distillation is proposed to achieve semantic knowledge transferring for segmentation. Extensive experiments on both synthetic and real-world multi-sensor benchmarks demonstrate that our method achieves superior performance.
format Preprint
id arxiv_https___arxiv_org_abs_2411_17141
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Learning Robust Anymodal Segmentor with Unimodal and Cross-modal Distillation
Zheng, Xu
Xue, Haiwei
Chen, Jialei
Yan, Yibo
Jiang, Lutao
Lyu, Yuanhuiyi
Yang, Kailun
Zhang, Linfeng
Hu, Xuming
Computer Vision and Pattern Recognition
Simultaneously using multimodal inputs from multiple sensors to train segmentors is intuitively advantageous but practically challenging. A key challenge is unimodal bias, where multimodal segmentors over rely on certain modalities, causing performance drops when others are missing, common in real world applications. To this end, we develop the first framework for learning robust segmentor that can handle any combinations of visual modalities. Specifically, we first introduce a parallel multimodal learning strategy for learning a strong teacher. The cross-modal and unimodal distillation is then achieved in the multi scale representation space by transferring the feature level knowledge from multimodal to anymodal segmentors, aiming at addressing the unimodal bias and avoiding over-reliance on specific modalities. Moreover, a prediction level modality agnostic semantic distillation is proposed to achieve semantic knowledge transferring for segmentation. Extensive experiments on both synthetic and real-world multi-sensor benchmarks demonstrate that our method achieves superior performance.
title Learning Robust Anymodal Segmentor with Unimodal and Cross-modal Distillation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2411.17141