Teachers in Concordance for Pseudo-Labeling of 3D Sequential Data

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F23%3A00362258" target="_blank" >RIV/68407700:21230/23:00362258 - isvavai.cz</a>

Výsledek na webu

<a href="https://doi.org/10.1109/LRA.2022.3226029" target="_blank" >https://doi.org/10.1109/LRA.2022.3226029</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1109/LRA.2022.3226029" target="_blank" >10.1109/LRA.2022.3226029</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Teachers in Concordance for Pseudo-Labeling of 3D Sequential Data

Popis výsledku v původním jazyce

Automatic pseudo-labeling is a powerful tool to tap into large amounts of sequential unlabeled data. It is especially appealing in safety-critical applications of autonomous driving, where performance requirements are extreme, datasets are large, and manual labeling is very challenging. We propose to leverage sequences of point clouds to boost the pseudo-labeling technique in a teacher-student setup via training multiple teachers, each with access to different temporal information. This set of teachers, dubbed Concordance , provides higher quality pseudo-labels for student training than standard methods. The output of multiple teachers is combined via a novel pseudo-label confidence-guided criterion. Our experimental evaluation focuses on the 3D point cloud domain and urban driving scenarios. We show the performance of our method applied to 3D semantic segmentation and 3D object detection on three benchmark datasets. Our approach, which uses only 20% manual labels, outperforms some fully supervised methods. A notable performance boost is achieved for classes rarely appearing in training data. Our codes will be made publicly available.

Název v anglickém jazyce

Teachers in Concordance for Pseudo-Labeling of 3D Sequential Data

Popis výsledku anglicky

Automatic pseudo-labeling is a powerful tool to tap into large amounts of sequential unlabeled data. It is especially appealing in safety-critical applications of autonomous driving, where performance requirements are extreme, datasets are large, and manual labeling is very challenging. We propose to leverage sequences of point clouds to boost the pseudo-labeling technique in a teacher-student setup via training multiple teachers, each with access to different temporal information. This set of teachers, dubbed Concordance , provides higher quality pseudo-labels for student training than standard methods. The output of multiple teachers is combined via a novel pseudo-label confidence-guided criterion. Our experimental evaluation focuses on the 3D point cloud domain and urban driving scenarios. We show the performance of our method applied to 3D semantic segmentation and 3D object detection on three benchmark datasets. Our approach, which uses only 20% manual labels, outperforms some fully supervised methods. A notable performance boost is achieved for classes rarely appearing in training data. Our codes will be made publicly available.

Druh

J_imp - Článek v periodiku v databázi Web of Science

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

Výsledek vznikl pri realizaci vícero projektů. Více informací v záložce Projekty.

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Rok uplatnění

2023

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název periodika

IEEE Robotics and Automation Letters

ISSN

2377-3766

e-ISSN

2377-3766

Svazek periodika

8

Číslo periodika v rámci svazku

2

Stát vydavatele periodika

US - Spojené státy americké

Počet stran výsledku

8

Strana od-do

536-543

Kód UT WoS článku

000902032700003

EID výsledku v databázi Scopus

2-s2.0-85144036267