AI Paper Atlas

Research map / Detection and segmentation / CNN features

Calibrated Adversarial Refinement for Stochastic Semantic Segmentation

Elias Kassapis, Georgi Dikov, Deepak K. Gupta, Cedric Nugteren

ICCV 2021 · Detection and segmentation › CNN features · approach: Context modeling for semantic segmentation · 18 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
replace calibration loss L_cal with cross entropy L_ce in the refinement network (ablation)CAR · LIDCGED (16 samples): 0.639 -> 0.264Improved
replace calibration loss L_cal with cross entropy L_ce in the refinement network (ablation)CAR · LIDCHM-IoU (16 samples): 0.477 -> 0.592Improved
replace calibration loss L_cal with cross entropy L_ce in the refinement network (ablation)CAR · Modified CityscapesGED (16 samples): 0.632 -> 0.164Improved
increase samples used for L_cal from 1 to 20 (ablation)CAR · LIDCGED (16 samples): 0.644 -> 0.264Improved
increase samples used for L_cal from 5 to 20 (ablation)CAR · LIDCGED (16 samples): 0.278 -> 0.264Improved
increase samples used for L_cal from 1 to 20 (ablation)CAR · LIDCHM-IoU (16 samples): 0.494 -> 0.592Improved
add the calibration loss to the regression GAN (ablation)GAN · 1D bimodal regressiondata log-likelihoodImproved
replace KL term with L_cal in cVAE-GAN (beta=0) (ablation)cVAE-GAN · LIDCGED (16 samples): 0.577 -> 0.272Improved
replace KL term with L_cal in cVAE-GAN (beta=0) versus beta=10 (ablation)cVAE-GAN · LIDCGED (16 samples): 0.609 -> 0.272Improved
replace KL term with L_cal in cVAE-GAN (beta=0) (ablation)cVAE-GAN · LIDCHM-IoU (16 samples): 0.484 -> 0.593Improved
use probabilistic encoder (cVAE-GAN+L_cal) instead of fixed noise (CAR) (ablation)CAR · LIDCGED (16 samples): 0.264 -> 0.272No clear effect
condition refinement network on ground-truth distribution instead of learned calibration network (ablation)CAR · Modified CityscapesGED (16 samples): 0.164 -> 0.038Improved

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