Image Classification on ImageNet-1K original (val)
80.1Top-1 AccScion
Evaluation results
| Method | Links | ||
|---|---|---|---|
| ScionBackbone=ViT-S/16, Training budget=300 epochs2025.12 | 80.1 | — | |
| ScionC (cosine)Backbone=ViT-S/16, Training budget=300 epochs2025.12 | 80.06 | — | |
| ScionC (constant)Backbone=ViT-S/16, Training budget=300 epochs2025.12 | 79.94 | — | |
| AdamWBackbone=ViT-S/16, Training budget=300 epochs2025.12 | 79.73 | — | |
| AdamCBackbone=ViT-S/16, Training budget=300 epochs2025.12 | 79.7 | — | |
| ScionBackbone=ViT-S/16, Training budget=150 epochs2025.12 | 79.65 | — | |
| ScionC (cosine)Backbone=ViT-S/16, Training budget=150 epochs2025.12 | 79.62 | — | |
| ScionC (constant)Backbone=ViT-S/16, Training budget=150 epochs2025.12 | 79.58 | — | |
| ScionC (cosine)Backbone=ViT-S/16, Training budget=90 epochs2025.12 | 78.74 | — | |
| AdamCBackbone=ViT-S/16, Training budget=150 epochs2025.12 | 78.69 | — | |
| ScionBackbone=ViT-S/16, Training budget=90 epochs2025.12 | 78.68 | — | |
| AdamWBackbone=ViT-S/16, Training budget=150 epochs2025.12 | 78.64 | — | |
| ScionC (constant)Backbone=ViT-S/16, Training budget=90 epochs2025.12 | 78.53 | — | |
| ScionBackbone=ViT-S/16, Training budget=60 epochs2025.12 | 77.44 | — | |
| ScionC (cosine)Backbone=ViT-S/16, Training budget=60 epochs2025.12 | 77.43 | — | |
| ScionC (constant)Backbone=ViT-S/16, Training budget=60 epochs2025.12 | 77.2 | — | |
| AdamCBackbone=ViT-S/16, Training budget=90 epochs2025.12 | 76.98 | — | |
| AdamWBackbone=ViT-S/16, Training budget=90 epochs2025.12 | 76.92 | — | |
| AdamWBackbone=ViT-S/16, Training budget=60 epochs2025.12 | 74.77 | — | |
| AdamCBackbone=ViT-S/16, Training budget=60 epochs2025.12 | 74.59 | — | |
| ScionBackbone=ViT-S/16, Training budget=30 epochs2025.12 | 73.31 | — | |
| ScionC (constant)Backbone=ViT-S/16, Training budget=30 epochs2025.12 | 73.1 | — | |
| ScionC (cosine)Backbone=ViT-S/16, Training budget=30 epochs2025.12 | 73.1 | — | |
| AdamCBackbone=ViT-S/16, Training budget=30 epochs2025.12 | 67.53 | — | |
| AdamWBackbone=ViT-S/16, Training budget=30 epochs2025.12 | 67.35 | — | |
| ConvNeXt-BSize (Px)=224, #Param (M)=88.6, FLOPs (G)=15.4, Throughput (Img/Sec)=14852023.06 | — | 83.8 | |
| ConvNeXt-LSize (Px)=224, #Param (M)=198.0, FLOPs (G)=34.4, Throughput (Img/Sec)=5082023.06 | — | 84.3 | |
| ConvNeXt-SSize (Px)=224, #Param (M)=50.2, FLOPs (G)=8.7, Throughput (Img/Sec)=20082023.06 | — | 83.1 | |
| ConvNeXt-TSize (Px)=224, #Param (M)=28.6, FLOPs (G)=4.5, Throughput (Img/Sec)=31962023.06 | — | 82 | |
| FasterViT-0Size (Px)=224, #Param (M)=31.4, FLOPs (G)=3.3, Throughput (Img/Sec)=58022023.06 | — | 82.1 | |
| FasterViT-1Size (Px)=224, #Param (M)=53.4, FLOPs (G)=5.3, Throughput (Img/Sec)=41882023.06 | — | 83.2 | |
| FasterViT-2Size (Px)=224, #Param (M)=75.9, FLOPs (G)=8.7, Throughput (Img/Sec)=31612023.06 | — | 84.2 | |
| FasterViT-3Size (Px)=224, #Param (M)=159.5, FLOPs (G)=18.2, Throughput (Img/Sec)=17802023.06 | — | 84.9 | |
| FasterViT-4Size (Px)=224, #Param (M)=424.6, FLOPs (G)=36.6, Throughput (Img/Sec)=8492023.06 | — | 85.4 | |
| FasterViT-5Size (Px)=224, #Param (M)=975.5, FLOPs (G)=113.0, Throughput (Img/Sec)=4492023.06 | — | 85.6 | |
| FasterViT-6Size (Px)=224, #Param (M)=1360.0, FLOPs (G)=142.0, Throughput (Img/Sec)=3522023.06 | — | 85.8 | |
| Swin-BSize (Px)=224, #Param (M)=87.8, FLOPs (G)=15.4, Throughput (Img/Sec)=12322023.06 | — | 83.4 | |
| Swin-SSize (Px)=224, #Param (M)=49.6, FLOPs (G)=8.5, Throughput (Img/Sec)=17202023.06 | — | 83.2 | |
| Swin-TSize (Px)=224, #Param (M)=28.3, FLOPs (G)=4.4, Throughput (Img/Sec)=27582023.06 | — | 81.3 |