List your company
ResearchFormerly WizWand

Image Classification on STL-10 OOD

98.6AccuracyLP-FT + VRF
78.5883.777588.97594.1725May 17, 2023Aug 15, 2023Nov 14, 2023Feb 13, 2024May 13, 2024Aug 12, 2024Nov 11, 2024
Updated 7mo ago

Evaluation results

MethodLinks
LP-FT + VRFBackbone=CLIP ViT-B/162024.1198.6
Zero-shotBackbone=CLIP ViT-B/162024.1198.4
E2E-FT + VRFBackbone=CLIP ViT-B/162024.1198.4
LP-FT + WSEBackbone=CLIP ViT-B/162024.1197.9
E2E-FT + WSEBackbone=CLIP ViT-B/162024.1197.8
Linear classifierBackbone=CLIP ViT-B/162024.1197.7
LP-FT + OSEBackbone=CLIP ViT-B/162024.1197.7
LP-FT + VRFBackbone=CLIP ViT-B/322024.1197.5
E2E-FT + VRFBackbone=CLIP ViT-B/322024.1197.3
Zero-shotBackbone=CLIP ViT-B/322024.1197.1
Linear classifierBackbone=CLIP ViT-B/322024.1196.6
E2E-FT + OSEBackbone=CLIP ViT-B/162024.1196.6
LP-FT + WSEBackbone=CLIP ViT-B/322024.1196.4
LP-FT + OSEBackbone=CLIP ViT-B/322024.1196.4
LP-FTBackbone=CLIP ViT-B/162024.1196.3
E2E-FTBackbone=CLIP ViT-B/162024.1196.1
E2E-FT + OSEBackbone=CLIP ViT-B/322024.1195.9
E2E-FT + WSEBackbone=CLIP ViT-B/322024.1195.7
LP-FTBackbone=CLIP ViT-B/322024.1195
E2E-FTBackbone=CLIP ViT-B/322024.1193.5
ResNeXt41-WMCGParams (M)=7.97, MACs (G)=10.05, MACs/Params (G/M)=1.28, Training Time (s/epoch)=31, Training VRAM Footprint (MB)=23184, Inference Time (s)=112023.0583.71
SESNParams (M)=10.96, MACs (G)=227.02, MACs/Params (G/M)=28.47, Training Time (s/epoch)=89, Training VRAM Footprint (MB)=30440, Inference Time (s)=482023.0582.3
ResNeXt41Params (M)=8.15, MACs (G)=10.05, MACs/Params (G/M)=1.26, Training Time (s/epoch)=27, Training VRAM Footprint (MB)=21886, Inference Time (s)=112023.0581.13
RST-CNNParams (M)=9.74, MACs (G)=850.23, MACs/Params (G/M)=106.62, Training Time (s/epoch)=437, Training VRAM Footprint (MB)=79248, Inference Time (s)=1842023.0579.35