AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

Latent Discriminant Deterministic Uncertainty

Gianni Franchi, Xuanlong Yu, Andrei Bursuc, Emanuel Aldea, Severine Dubuisson, David Filliat

ECCV 2022 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Priors and functional variational uncertainty · 14 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add entropy loss to uncertainty loss (ablation)LDU · CIFAR-10 OOD vs SVHNAUC: 0.8195 -> 0.8362Improved
add discriminative loss to uncertainty loss (ablation)LDU · CIFAR-10 OOD vs SVHNAUC: 0.8195 -> 0.8613Improved
add all three losses versus uncertainty loss alone (ablation)LDU · CIFAR-10 OOD vs SVHNAUC: 0.8195 -> 0.8721Improved
add entropy loss to uncertainty plus discriminative loss (ablation)LDU · CIFAR-10 OOD vs SVHNAUC: 0.8613 -> 0.8721Improved
add entropy loss to uncertainty loss (ablation)LDU · CIFAR-10ECE: 0.5256 -> 0.5386Hurt
increase prototypes from 16 to 128 (ablation)LDU · CIFAR-10 OOD vs SVHNAUC: 0.8479 -> 0.8721Mixed
raise lambda from 0.1 to 2.0 (ablation)LDU · CIFAR-10 OOD vs SVHNAUC: 0.8721 -> 0.8339Hurt
use LDU instead of baseline MCPDNN · CIFAR-10 OOD vs SVHNAUC: 0.8032 -> 0.8721Improved
use LDU instead of baseline MCPDeeplab v3+ · MUAD OOD setAUPR: 0.2604 -> 0.4418Improved
use LDU instead of Deep EnsemblesDeeplab v3+ · MUAD OOD setAUROC: 0.871 -> 0.8816Improved
use LDU instead of Deep EnsemblesDeeplab v3+ · CityscapesmIoU: 77.23% -> 76.62%No clear effect
use LDU instead of Deep EnsemblesDeeplab v3+ · semantic segmentation inference timeruntime ms: 56 -> 14Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy