Image Classification on ImageNet-1k (Accuracy, GFLOPs, Throughput)
78.55AccuracyViT-L
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ViT-LParams.=304.35M, Transformation=Identity2024.10 | 78.55 | 63.54 | 1,219.8 | |
| ViT-LParams.=304.35M, Transformation=Linear2024.10 | 78.55 | 63.54 | 1,225.2 | |
| TOASTBackbone=ViT-L, Approx.=3 → 4, Params.=-11.55M, Transformation=Linear2024.10 | 78.36 | 61.11 | 1,270 | |
| TOASTBackbone=ViT-L, Approx.=22 → 23, Params.=-11.55M, Transformation=Identity2024.10 | 78.32 | 60.9 | 1,267.5 | |
| TOASTBackbone=ViT-L, Approx.=22 → 23, Params.=-11.55M, Transformation=Linear2024.10 | 78.21 | 61.11 | 1,270.9 | |
| TOASTBackbone=ViT-L, Approx.=16 → 17, 22 → 23, Params.=-23.09M, Transformation=Identity2024.10 | 77.64 | 58.26 | 1,318.8 | |
| TOASTBackbone=ViT-L, Approx.=16 → 17, 22 → 23, Params.=-23.09M, Transformation=Linear2024.10 | 77.64 | 58.67 | 1,312.3 | |
| TOASTBackbone=ViT-L, Approx.=7 → 8, 15 → 16, Params.=-23.09M, Transformation=Linear2024.10 | 77.32 | 58.67 | 1,316.4 | |
| TOASTBackbone=ViT-L, Approx.=3 → 4, Params.=-11.55M, Transformation=Identity2024.10 | 77.32 | 60.9 | 1,269.2 | |
| TOASTBackbone=ViT-L, Approx.=7 → 8, 15 → 16, Params.=-23.09M, Transformation=Identity2024.10 | 76.81 | 58.26 | 1,318.2 | |
| TOASTBackbone=ViT-L, Approx.=3 → 4, 21 → 23, Params.=-34.64M, Transformation=Linear2024.10 | 76.3 | 56.03 | 1,345.6 | |
| TOASTBackbone=ViT-L, Approx.=3 → 4, 21 → 23, Params.=-34.64M, Transformation=Identity2024.10 | 75.8 | 55.62 | 1,377.2 | |
| DINO-BParams.=86.58M, Transformation=Identity2024.10 | 74.52 | 24.42 | 2,321.3 | |
| DINO-BParams.=86.58M, Transformation=Linear2024.10 | 74.52 | 24.42 | 2,316.5 | |
| TOASTBackbone=ViT-L, Approx.=3 → 4, 20 → 23, Params.=-46.19M, Transformation=Linear2024.10 | 74.03 | 53.39 | 1,436.8 | |
| DeiT-SParams.=21.81M, Transformation=Identity2024.10 | 73.85 | 4.97 | 6,349.2 | |
| DeiT-SParams.=21.81M, Transformation=Linear2024.10 | 73.85 | 4.97 | 6,325.6 | |
| TOASTBackbone=DeiT-S, Approx.=10 → 11, Params.=-1.62M, Transformation=Linear2024.10 | 73.78 | 4.59 | 6,740.5 | |
| TOASTBackbone=DeiT-S, Approx.=10 → 11, Params.=-1.62M, Transformation=Identity2024.10 | 73.67 | 4.56 | 6,752.6 | |
| TOASTBackbone=ViT-L, Approx.=3 → 4, 20 → 23, Params.=-46.19M, Transformation=Identity2024.10 | 73.49 | 52.98 | 1,440.4 | |
| TOASTBackbone=DINO-B, Approx.=0 → 1, Params.=-6.5M, Transformation=Linear2024.10 | 73.43 | 22.54 | 2,487 | |
| TOASTBackbone=DeiT-S, Approx.=2 → 3, Params.=-1.62M, Transformation=Linear2024.10 | 73.19 | 4.59 | 6,736.7 | |
| TOASTBackbone=DINO-B, Approx.=5 → 6, Params.=-6.5M, Transformation=Linear2024.10 | 73.01 | 22.54 | 2,490.6 | |
| TOASTBackbone=ViT-L, Approx.=3 → 4, 19 → 23, Params.=-57.74M, Transformation=Linear2024.10 | 71.5 | 50.75 | 1,499.5 | |
| TOASTBackbone=DeiT-S, Approx.=3 → 4, 9 → 10, Params.=-3.25M, Transformation=Linear2024.10 | 71.35 | 4.21 | 7,188.4 | |
| TOASTBackbone=ViT-L, Approx.=3 → 4, 19 → 23, Params.=-57.74M, Transformation=Identity2024.10 | 70.97 | 50.34 | 1,509.9 | |
| TOASTBackbone=DINO-B, Approx.=0 → 1, 2 → 3, Params.=-13M, Transformation=Linear2024.10 | 70.82 | 20.67 | 2,681.2 | |
| TOASTBackbone=DeiT-S, Approx.=2 → 3, Params.=-1.62M, Transformation=Identity2024.10 | 70.8 | 4.56 | 6,754.2 | |
| TOASTBackbone=DeiT-S, Approx.=3 → 4, 9 → 10, Params.=-3.25M, Transformation=Identity2024.10 | 69.22 | 4.15 | 7,210.9 | |
| TOASTBackbone=DeiT-S, Approx.=3 → 4, 9 → 11, Params.=-4.88M, Transformation=Linear2024.10 | 68.39 | 3.8 | 7,718.4 | |
| TOASTBackbone=ViT-L, Approx.=2 → 4, 18 → 23, Params.=-80.83M, Transformation=Linear2024.10 | 67.43 | 45.47 | 1,652.8 | |
| TOASTBackbone=DeiT-S, Approx.=3 → 4, 9 → 11, Params.=-4.88M, Transformation=Identity2024.10 | 66.96 | 3.74 | 7,751.4 | |
| TOASTBackbone=ViT-L, Approx.=17 → 23, Params.=-69.28M, Transformation=Linear2024.10 | 66.87 | 47.9 | 1,567 | |
| TOASTBackbone=ViT-L, Approx.=17 → 23, Params.=-69.28M, Transformation=Identity2024.10 | 66.81 | 47.7 | 1,572.4 | |
| TOASTBackbone=DINO-B, Approx.=0 → 1, Params.=-6.5M, Transformation=Identity2024.10 | 65.47 | 22.39 | 2,506.6 | |
| TOASTBackbone=ViT-L, Approx.=2 → 4, 18 → 23, Params.=-80.83M, Transformation=Identity2024.10 | 62.92 | 45.05 | 1,654.9 | |
| TOASTBackbone=DINO-B, Approx.=0 → 1, 2 → 3, 4 → 5, Params.=-19.5M, Transformation=Linear2024.10 | 61.02 | 18.8 | 2,929.6 | |
| TOASTBackbone=DINO-B, Approx.=5 → 6, Params.=-6.5M, Transformation=Identity2024.10 | 28.84 | 22.39 | 2,503.1 | |
| TOASTBackbone=DINO-B, Approx.=0 → 4, Params.=-26M, Transformation=Linear2024.10 | 27.7 | 16.47 | 3,227.7 | |
| TOASTBackbone=DINO-B, Approx.=0 → 1, 2 → 3, Params.=-13M, Transformation=Identity2024.10 | 13.28 | 20.37 | 2,703.9 | |
| TOASTBackbone=DINO-B, Approx.=0 → 1, 2 → 3, 4 → 5, Params.=-19.5M, Transformation=Identity2024.10 | 6.98 | 18.34 | 2,947 | |
| TOASTBackbone=DINO-B, Approx.=0 → 4, Params.=-26M, Transformation=Identity2024.10 | 3.58 | 16.32 | 3,230.9 |