Explainable AI (XAI) Inference Efficiency on ImageNet (1000 images)
7Inference Time (s)FEX
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| FEXBackbone=pretrained ViT model, Hardware=8 CPU cores and 1 Nvidia A100 GPU, Sample Size=1000 image predictions2024.05 | 7 | 2 | 14 | |
| FastSHAPBackbone=pretrained ViT model, Hardware=8 CPU cores and 1 Nvidia A100 GPU, Sample Size=1000 image predictions2024.05 | 11.6 | 1.2 | 13.9 | |
| GradCAMBackbone=pretrained ViT model, Hardware=8 CPU cores and 1 Nvidia A100 GPU, Sample Size=1000 image predictions2024.05 | 14.9 | 1.9 | 28.3 | |
| AttLRPBackbone=pretrained ViT model, Hardware=8 CPU cores and 1 Nvidia A100 GPU, Sample Size=1000 image predictions2024.05 | 106.8 | 2 | 213.6 | |
| RISEBackbone=pretrained ViT model, Hardware=8 CPU cores and 1 Nvidia A100 GPU, Sample Size=1000 image predictions2024.05 | 260.2 | 15.9 | 4,137.2 | |
| IGBackbone=pretrained ViT model, Hardware=8 CPU cores and 1 Nvidia A100 GPU, Sample Size=1000 image predictions2024.05 | 311.9 | 24.5 | 7,641.6 | |
| GradSHAPBackbone=pretrained ViT model, Hardware=8 CPU cores and 1 Nvidia A100 GPU, Sample Size=1000 image predictions2024.05 | 313.2 | 7.1 | 2,223.7 |