Portrait de Sukesh Adiga Vasudeva

Sukesh Adiga Vasudeva

Doctorat - Polytechnique
Superviseur⋅e principal⋅e
Hervé Lombaert
Sujets de recherche
Apprentissage automatique médical
Apprentissage de représentations
Apprentissage profond
Optimisation
Segmentation d'images médicales
Vision par ordinateur
Site web
Google Scholar
GitHub

Publications

GeoLS: an Intensity-based, Geodesic Soft Labeling for Image Segmentation
Sukesh Adiga Vasudeva
Jose Dolz
Hervé Lombaert
2025-04-18
Machine Learning for Biomedical Imaging (publié)
doi.org
Neighbor-Aware Calibration of Segmentation Networks with Penalty-Based Constraints
Balamurali Murugesan
Sukesh Adiga Vasudeva
Bingyuan Liu
Hervé Lombaert
Ismail Ben Ayed
Jose Dolz
Ensuring reliable confidence scores from deep neural networks is of paramount significance in critical decision-making systems, particularly… (voir plus) in real-world domains such as healthcare. Recent literature on calibrating deep segmentation networks has resulted in substantial progress. Nevertheless, these approaches are strongly inspired by the advancements in classification tasks, and thus their uncertainty is usually modeled by leveraging the information of individual pixels, disregarding the local structure of the object of interest. Indeed, only the recent Spatially Varying Label Smoothing (SVLS) approach considers pixel spatial relationships across classes, by softening the pixel label assignments with a discrete spatial Gaussian kernel. In this work, we first present a constrained optimization perspective of SVLS and demonstrate that it enforces an implicit constraint on soft class proportions of surrounding pixels. Furthermore, our analysis shows that SVLS lacks a mechanism to balance the contribution of the constraint with the primary objective, potentially hindering the optimization process. Based on these observations, we propose NACL (Neighbor Aware CaLibration), a principled and simple solution based on equality constraints on the logit values, which enables to control explicitly both the enforced constraint and the weight of the penalty, offering more flexibility. Comprehensive experiments on a wide variety of well-known segmentation benchmarks demonstrate the superior calibration performance of the proposed approach, without affecting its discriminative power. Furthermore, ablation studies empirically show the model agnostic nature of our approach, which can be used to train a wide span of deep segmentation networks.
2024-01-25
ArXiv (prépublication)
doi.org
arxiv.org