List your company
ResearchFormerly WizWand

Surface Normal Estimation on PASCAL-Context (test)

13.44mErrUV-M3TL
13.246814.550915.85517.1591Nov 22, 2023Apr 3, 2024Aug 15, 2024Dec 27, 2024May 10, 2025Sep 21, 2025Feb 2, 2026
Updated 8mo ago

Evaluation results

MethodLinks
UV-M3TLBackbone=ViT-large2026.0213.44——
MLOREBackbone=ViT-large2024.0313.51——
MLoREBackbone=ViT-large2026.0213.51——
TaskDiffusionBackbone=ViT-large2026.0213.55——
TaskExpertBackbone=ViT-large2024.0313.56——
TaskExpertBackbone=ViT-large2026.0213.56——
TSP-TransformerBackbone=ViT-large2026.0213.69——
Mod-SquadFLOPs (G)=420, #Param (M)=522024.0313.7——
TaskPrompterBackbone=ViT-large2024.0313.72——
TaskPrompterBackbone=ViT-large2026.0213.72——
InvPT++Backbone=ViT-large2026.0213.73——
FEDHCA2Client Setup=Multi-Task2023.1114.080.75—
MTMambaBackbone=Swin-Large2026.0214.14——
InvPTBackbone=ViT-large2024.0314.15——
InvPTBackbone=ViT-large2026.0214.15——
FedAMPClient Setup=Multi-Task2023.1114.20.27—
ATRCBackbone=HRNet182026.0214.2——
LocalClient Setup=Multi-Task2023.1114.210—
MQTransformerBackbone=not specified2026.0214.21——
TaskExpertFLOPs (G)=204, #Param (M)=552024.0314.22——
MTI-NetBackbone=HRNet182024.0314.23——
ATRCBackbone=ViT-large, reproduced=true2024.0314.23——
MTI-NetBackbone=HRNet182026.0214.23——
ATRC (Reproduced)Backbone=ViT-large2026.0214.23——
ATRCBackbone=HRNet182024.0314.24——
ATRC-BMTASBackbone=HRNet182026.0214.24——
MTMamba++Backbone=Swin-Large2026.0214.29——
ATRC-ASPPBackbone=HRNet182026.0214.3——
PAD-NetBackbone=ViT-large, reproduced=true2024.0314.37——
PAD-Net (Reproduced)Backbone=ViT-large2026.0214.37——
MLOREFLOPs (G)=72, #Param (M)=442024.0314.43——
M³ViTFLOPs (G)=420, #Param (M)=422024.0314.5——
MaT-FLClient Setup=Multi-Task2023.1114.58-0.16—
MTI-NetBackbone=ViT-large, reproduced=true2024.0314.67——
MTI-Net (Reproduced)Backbone=ViT-large2026.0214.67——
ASTMTBackbone=not specified2026.0214.7——
FedPerClient Setup=Multi-Task2023.1114.75-0.16—
DittoClient Setup=Multi-Task2023.1115.29-4.67—
PAD-NetBackbone=HRNet182024.0315.3——
PAD-NetBackbone=HRNet182026.0215.3——
FedAvgClient Setup=Multi-Task2023.1115.56-7.56—
FEDHCA^2Backbone=Swin-T, Client Setup=Single-Task2023.1115.6——
FedAMPBackbone=Swin-T, Client Setup=Single-Task2023.1115.74——
LocalBackbone=Swin-T, Client Setup=Single-Task2023.1115.76——
FedProxClient Setup=Multi-Task2023.1115.83-8.12—
FedPerBackbone=Swin-T, Client Setup=Single-Task2023.1116.95——
MaT-FLBackbone=Swin-T, Client Setup=Single-Task2023.1117.26——
DittoBackbone=Swin-T, Client Setup=Single-Task2023.1117.52——
FedProxBackbone=Swin-T, Client Setup=Single-Task2023.1118.03——
FedAvgBackbone=Swin-T, Client Setup=Single-Task2023.1118.27——
AdapterBackbone=Swin-Tiny, Trainable Parameters (M)=11.24, Single Inference For All Tasks=false2024.03——18.83
BitfitBackbone=Swin-Tiny, Trainable Parameters (M)=2.85, Single Inference For All Tasks=false2024.03——19.42
CompactorBackbone=Swin-Tiny, Trainable Parameters (M)=2.78, Single Inference For All Tasks=false2024.03——19.22
Compactor++Backbone=Swin-Tiny, Trainable Parameters (M)=2.66, Single Inference For All Tasks=false2024.03——19.54
HyperFormerBackbone=Swin-Tiny, Trainable Parameters (M)=72.77, Single Inference For All Tasks=false2024.03——17.77
LoRABackbone=Swin-Tiny, Trainable Parameters (M)=2.87, Single Inference For All Tasks=false2024.03——18.96
MTL - Full Fine TuningBackbone=Swin-Tiny, Trainable Parameters (M)=30.06, Single Inference For All Tasks=true2024.03——16.64
MTL - Tuning Decoders OnlyBackbone=Swin-Tiny, Trainable Parameters (M)=1.94, Single Inference For All Tasks=true2024.03——20.69
MTLORABackbone=Swin-Tiny, Rank (r)=16, Trainable Parameters (M)=4.95, Single Inference For All Tasks=true2024.03——17.03
MTLORABackbone=Swin-Tiny, Rank (r)=32, Trainable Parameters (M)=6.08, Single Inference For All Tasks=true2024.03——16.59
MTLORABackbone=Swin-Tiny, Rank (r)=64, Trainable Parameters (M)=8.34, Single Inference For All Tasks=true2024.03——16.6
MTLORA+Backbone=Swin-Tiny, Rank (r)=4, Trainable Parameters (M)=2.57, Single Inference For All Tasks=true2024.03——17.6
MTLORA+Backbone=Swin-Tiny, Rank (r)=8, Trainable Parameters (M)=3.15, Single Inference For All Tasks=true2024.03——17.41
MTLORA+Backbone=Swin-Tiny, Rank (r)=16, Trainable Parameters (M)=4.29, Single Inference For All Tasks=true2024.03——17.03
PolyhistorBackbone=Swin-Tiny, Trainable Parameters (M)=8.96, Single Inference For All Tasks=false2024.03——17.77
Single TaskBackbone=Swin-Tiny, Trainable Parameters (M)=112.62, Single Inference For All Tasks=false2024.03——17.97
VL-AdapterBackbone=Swin-Tiny, Trainable Parameters (M)=4.74, Single Inference For All Tasks=false2024.03——19.26
VPT-deepBackbone=Swin-Tiny, Trainable Parameters (M)=3.43, Single Inference For All Tasks=false2024.03——21.07
VPT-shallowBackbone=Swin-Tiny, Trainable Parameters (M)=2.57, Single Inference For All Tasks=false2024.03——20.9