3D LiDAR semantic segmentation is fundamental for autonomous driving. Several Unsupervised Domain Adaptation (UDA) methods for point cloud data have been recently proposed to improve model generalization for different sensors and environments. Researchers working on UDA problems in the image domain have shown that sample mixing can mitigate domain shift. We propose a new approach of sample mixing for point cloud UDA, namely Compositional Semantic Mix (CoSMix), the first UDA approach for point cloud segmentation based on sample mixing. CoSMix consists of a two-branch symmetric network that can process labelled synthetic data (source) and real-world unlabelled point clouds (target) concurrently. Each branch operates on one domain by mixing selected pieces of data from the other one, and by using the semantic information derived from source labels and target pseudo-labels. We evaluate CoSMix on two large-scale datasets, showing that it outperforms state-of-the-art methods by a large margin.

CoSMix: Compositional Semantic Mix for Domain Adaptation in 3D LiDAR Segmentation / Saltori, Cristiano; Galasso, Fabio; Fiameni, Giuseppe; Sebe, Nicu; Ricci, Elisa; Poiesi, Fabio. - 13693:(2022), pp. 586-602. (Intervento presentato al convegno 17th European Conference on Computer Vision, ECCV 2022 tenutosi a Tel Aviv nel 2022) [10.1007/978-3-031-19827-4_34].

CoSMix: Compositional Semantic Mix for Domain Adaptation in 3D LiDAR Segmentation

Saltori, Cristiano
;
Sebe, Nicu;Ricci, Elisa;Poiesi, Fabio
2022-01-01

Abstract

3D LiDAR semantic segmentation is fundamental for autonomous driving. Several Unsupervised Domain Adaptation (UDA) methods for point cloud data have been recently proposed to improve model generalization for different sensors and environments. Researchers working on UDA problems in the image domain have shown that sample mixing can mitigate domain shift. We propose a new approach of sample mixing for point cloud UDA, namely Compositional Semantic Mix (CoSMix), the first UDA approach for point cloud segmentation based on sample mixing. CoSMix consists of a two-branch symmetric network that can process labelled synthetic data (source) and real-world unlabelled point clouds (target) concurrently. Each branch operates on one domain by mixing selected pieces of data from the other one, and by using the semantic information derived from source labels and target pseudo-labels. We evaluate CoSMix on two large-scale datasets, showing that it outperforms state-of-the-art methods by a large margin.
2022
European Conference on Computer Vision (ECCV)
Heidelberg
Springer
978-3-031-19826-7
978-3-031-19827-4
Saltori, Cristiano; Galasso, Fabio; Fiameni, Giuseppe; Sebe, Nicu; Ricci, Elisa; Poiesi, Fabio
CoSMix: Compositional Semantic Mix for Domain Adaptation in 3D LiDAR Segmentation / Saltori, Cristiano; Galasso, Fabio; Fiameni, Giuseppe; Sebe, Nicu; Ricci, Elisa; Poiesi, Fabio. - 13693:(2022), pp. 586-602. (Intervento presentato al convegno 17th European Conference on Computer Vision, ECCV 2022 tenutosi a Tel Aviv nel 2022) [10.1007/978-3-031-19827-4_34].
File in questo prodotto:
File Dimensione Formato  
136930575.pdf

Open Access dal 03/11/2023

Tipologia: Post-print referato (Refereed author’s manuscript)
Licenza: Tutti i diritti riservati (All rights reserved)
Dimensione 9.45 MB
Formato Adobe PDF
9.45 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11572/361312
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 28
  • ???jsp.display-item.citation.isi??? 17
  • OpenAlex ND
social impact