Over-MAP: Structural Attention Mechanism and Automated Semantic Segmentation Ensembled for Uncertainty Prediction - Archive ouverte HAL
Communication Dans Un Congrès Année : 2021

Over-MAP: Structural Attention Mechanism and Automated Semantic Segmentation Ensembled for Uncertainty Prediction

Résumé

Both theoretical and practical problems in deep learning classification require solutions for assessing uncertainty prediction but current state-of-the-art methods in this area are computationally expensive. In this paper, we propose a new confidence measure dubbed Over-MAP that utilizes a measure of overlap between structural attention mechanisms and segmentation methods, that is of particular interest in accurate fine-grained contexts. We show that this classification confidence increases with the degree of overlap. The associated confidence and identification tools are conceptually simple, efficient, and of high practical interest as they allow for weeding out misleading examples in training data. Our measure is currently deployed in the real-world on widely used platforms to annotate large-scale data efficiently.
Fichier principal
Vignette du fichier
17798-Article Text-21292-1-2-20210518.pdf (1.42 Mo) Télécharger le fichier
Origine Fichiers éditeurs autorisés sur une archive ouverte

Dates et versions

hal-03541009 , version 1 (24-01-2022)

Identifiants

  • HAL Id : hal-03541009 , version 1

Citer

Charles A Kantor, Léonard Boussioux, Brice Rauby, Hugues Talbot. Over-MAP: Structural Attention Mechanism and Automated Semantic Segmentation Ensembled for Uncertainty Prediction. AAAI 2021 - 35th AAAI Conference on Artificial Intelligence, Feb 2021, Virtual, China. pp.15316-15322. ⟨hal-03541009⟩
134 Consultations
121 Téléchargements

Partager

More