ALLNet: Multi-task Dense Prediction for Degraded Images
Weiran Wang, Jialing Wu, Yaqi Chang, Gang He, Li Xu, Chang Wu, Yunsong Li
Abstract
Multi-taskdensepredictionaims tosimultaneously addressmultiplepixel-level tasksthroughaunifiednetworkforvisualsceneunderstanding.However,adverse environmental conditions limit thegeneralizationand practicality of such tasks. Toaddress this, we proposeALLNet, anovel framework that effectively explores degradationpatterns and integratesmulti-task collaborative information. Specifically, we designa MoE-basedMixtureofAdaptiveExperts(MaE)restorationcomponentnetworkthatenhancesdegradationfeaturesthroughdynamicroutingandguidestask-specific featureextraction. Furthermore,weformulateaTaskawareCollaborativeRefinement (TCR)moduletocaptureglobalsemanticcorrelationsandcross-taskdependencies,facilitatingbidirectionalcollaborationbetween restorationandtask-specific featuresondegradedimages. Tothebestofourknowledge, thisrepresentsthe firstattemptatmulti-taskdensepredictionunderimage degradation.ExperimentalresultsondegradedNYUDv2andPASCAL-Contextbenchmarksdemonstratethat ourarchitecturesignificantlyoutperformsexistingmethodsindegradedscenarios.
BibTeX
@inproceedings{cvpr2026_allnetmultitaskd,
title = {ALLNet: Multi-task Dense Prediction for Degraded Images},
author = {Weiran Wang and Jialing Wu and Yaqi Chang and Gang He and Li Xu and Chang Wu and Yunsong Li},
booktitle = {CVPR 2026},
year = {2026}
}