Thesis Details
Hodnocení neurčitosti predikcí neuronových sítí v úlohách klasifikace, detekce a segmentace
This work focuses on comparing three widely used methods for improving uncertainty estimations: Deep Ensembles, Monte Carlo Dropout, and Temperature Scaling. These methods are applied to six computer vision models that are pretrained as well as trained fromscratch. The models are then evaluated on computer vision datasets for classification, semantic segmentation, and object detection using a wide range of metrics. The models are also evaluated on distorted versions of these datasets to measure their performance on out-of-distribution data. These modified models achieve promising results. Ensembles outperform the other models by as high as 70 % in accuracy and 0.2 in IOU on the distorted MedSeg COVID-19 segmentation dataset while also outperforming the other models on the CIFAR-100 and FMNIST datasets.
monte carlo dropout, deep ensembles, temperature scaling, classification, semantic segmentation, object detection, dataset shift, model calibration, FMNIST, CIFAR-100, PASCAL-VOC
Beran Vítězslav, doc. Ing., Ph.D. (DCGM FIT BUT), člen
Burget Radek, doc. Ing., Ph.D. (DIFS FIT BUT), člen
Fučík Otto, doc. Dr. Ing. (DCSY FIT BUT), člen
Holík Lukáš, doc. Mgr., Ph.D. (DITS FIT BUT), člen
@bachelorsthesis{FITBT25039,
author = "Ji\v{r}\'{i} Vlas\'{a}k",
type = "Bachelor's thesis",
title = "Hodnocen\'{i} neur\v{c}itosti predikc\'{i} neuronov\'{y}ch s\'{i}t\'{i} v \'{u}loh\'{a}ch klasifikace, detekce a segmentace",
school = "Brno University of Technology, Faculty of Information Technology",
year = 2022,
location = "Brno, CZ",
language = "czech",
url = "https://www.fit.vut.cz/study/thesis/25039/"
}