List your company
ResearchFormerly WizWand

Image Classification on ImageNet-Sketch (val)

67.3Top-1 AccEVA CLIP-g
-1.8616.09534.0552.005Oct 27, 2021May 23, 2022Dec 17, 2022Jul 13, 2023Feb 6, 2024Sep 1, 2024Mar 29, 2025
Updated 8mo ago

Evaluation results

MethodLinks
EVA CLIP-gzero-shot evaluation protocol=True2022.1167.3
Open CLIP-Hzero-shot evaluation protocol=True2022.1166.6
Open CLIP-gzero-shot evaluation protocol=True2022.1165.2
ALIGNzero-shot evaluation protocol=True2022.1164.8
PaLI-17BShot Setting=0-shot2022.0963.83
PaLI-3BShot Setting=0-shot2022.0961.11
PaLI-15BShot Setting=0-shot2022.0961.03
OpenAI CLIP-Lzero-shot evaluation protocol=True2022.1159.6
MAEModel=ViT-H, Resolution=448, Training Objective=MAE pre-training2021.1150.9
MAEModel=ViT-H, Resolution=224, Training Objective=MAE pre-training2021.1149.6
OTTEREvaluation Protocol=Linear Probing, Label Distribution Estimation=BBSE+OT2024.0448.3
MAEModel=ViT-L, Resolution=224, Training Objective=MAE pre-training2021.1145.3
Linear Probing baselineEvaluation Protocol=Linear Probing, Label Distribution Estimation=None2024.0443.4
OTTEREvaluation Protocol=Zero-Shot, Label Distribution Estimation=BBSE+OT2024.0440.4
Zero-Shot baselineEvaluation Protocol=Zero-Shot, Label Distribution Estimation=None2024.0439.8
Supervised BaselineModel=ViT-H, Resolution=224, Training Objective=Supervised2021.1138
BBSE+PMEvaluation Protocol=Linear Probing, Label Distribution Estimation=BBSE+PM2024.0437.9
Supervised BaselineModel=ViT-L, Resolution=224, Training Objective=Supervised2021.1137.5
MSNArchitecture=ViT-B/16, Pre-training method=MSN2022.0436.3
LPTBackbone=ViT-B2022.1036.22
Noisy StudentNotes=Previous Best2021.1136
Supervised BaselineModel=ViT-B, Resolution=224, Training Objective=Supervised2021.1135.6
MAEModel=ViT-B, Resolution=224, Training Objective=MAE pre-training2021.1134.5
MAEArchitecture=ViT-B/16, Pre-training method=MAE2022.0434.5
Full fine-tuneBackbone=ViT-B2022.1032.25
Linear ProbeBackbone=ViT-B2022.1031.55
LSNet-BFLOPS=1.3G2025.0330.7
PVTv2-B1FLOPS=2.1G2025.0328.9
EdgeNeXt-SFLOPS=1.3G2025.0328.8
FastViT-T12FLOPS=1.4G2025.0327.6
LSNet-SFLOPS=0.5G2025.0327.5
FasterNet-T2FLOPS=1.9G2025.0327.2
UniRepLKNet-AFLOPS=0.6G2025.0326
LSNet-TFLOPS=0.3G2025.0325.5
FastViT-T8FLOPS=0.7G2025.0325.5
PoolFormer-S12FLOPS=1.8G2025.0325.2
ResNet50Architecture=ResNet50, Pre-training method=Supervised2022.0424.2
EfficientViT-M3FLOPS=0.3G2025.0323.4
EdgeNeXt-XSFLOPS=0.5G2025.0322
StarNet-S1FLOPS=0.4G2025.0321.8
PVTv2-B0FLOPS=0.6G2025.0321.5
PVT-TinyFLOPS=1.9G2025.0321.5
MoCo-v2 + Patch-base NSalpha=2, k=163842021.1019.49
MoCo-v2 + Patch-base NSalpha=3, k=655362021.1018.79
MoCo-v2 + Patch-base NSalpha=2, k=327682021.1018.7
MoCo-v2 + Patch-base NSalpha=2, k=655362021.1018.58
EdgeNeXt-XXSFLOPS=0.3G2025.0318.5
MoCo-v22021.1017.47
MoCo-v2 + MoCHiN=512, K=1024, s=5122021.1016.32
FasterNet-T0FLOPS=0.3G2025.0316.3
MoCo-v2 + MoCHiN=128, K=1024, s=5122021.1016
BBSE+PMEvaluation Protocol=Zero-Shot, Label Distribution Estimation=BBSE+PM2024.040.8