Enregistré dans:
Détails bibliographiques
Auteurs principaux: Wu, Fei, Marquez-Neila, Pablo, Rafi-Tarii, Hedyeh, Sznitman, Raphael
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:https://arxiv.org/abs/2412.06470
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866913737327247360
author Wu, Fei
Marquez-Neila, Pablo
Rafi-Tarii, Hedyeh
Sznitman, Raphael
author_facet Wu, Fei
Marquez-Neila, Pablo
Rafi-Tarii, Hedyeh
Sznitman, Raphael
contents Multi-class semantic segmentation remains a cornerstone challenge in computer vision. Yet, dataset creation remains excessively demanding in time and effort, especially for specialized domains. Active Learning (AL) mitigates this challenge by selecting data points for annotation strategically. However, existing patch-based AL methods often overlook boundary pixels critical information, essential for accurate segmentation. We present OREAL, a novel patch-based AL method designed for multi-class semantic segmentation. OREAL enhances boundary detection by employing maximum aggregation of pixel-wise uncertainty scores. Additionally, we introduce one-vs-rest entropy, a novel uncertainty score function that computes class-wise uncertainties while achieving implicit class balancing during dataset creation. Comprehensive experiments across diverse datasets and model architectures validate our hypothesis.
format Preprint
id arxiv_https___arxiv_org_abs_2412_06470
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Active Learning with Context Sampling and One-vs-Rest Entropy for Semantic Segmentation
Wu, Fei
Marquez-Neila, Pablo
Rafi-Tarii, Hedyeh
Sznitman, Raphael
Computer Vision and Pattern Recognition
Machine Learning
Multi-class semantic segmentation remains a cornerstone challenge in computer vision. Yet, dataset creation remains excessively demanding in time and effort, especially for specialized domains. Active Learning (AL) mitigates this challenge by selecting data points for annotation strategically. However, existing patch-based AL methods often overlook boundary pixels critical information, essential for accurate segmentation. We present OREAL, a novel patch-based AL method designed for multi-class semantic segmentation. OREAL enhances boundary detection by employing maximum aggregation of pixel-wise uncertainty scores. Additionally, we introduce one-vs-rest entropy, a novel uncertainty score function that computes class-wise uncertainties while achieving implicit class balancing during dataset creation. Comprehensive experiments across diverse datasets and model architectures validate our hypothesis.
title Active Learning with Context Sampling and One-vs-Rest Entropy for Semantic Segmentation
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2412.06470