Discovering Causes of Incorrect Predictions on CUB-200 2011 (val)
0.5339Avg Highest Confidence (Top 25%)LIMA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| LIMABackbone=MobileNetV2, Sub-region division=Segment Anything2025.04 | 0.5339 | 0.6098 | 0.6611 | 0.7349 | 0.3297 | |
| LIMABackbone=EfficientNetV2-M, Sub-region division=Segment Anything2025.04 | 0.4138 | 0.5035 | 0.5543 | 0.5999 | 0.3857 | |
| LIMABackbone=ResNet-101, Sub-region division=Segment Anything2025.04 | 0.3874 | 0.4957 | 0.5975 | 0.7916 | 0.403 | |
| LIMABackbone=ResNet-101, Sub-region division=SLICO2025.04 | 0.3683 | 0.5016 | 0.5501 | 0.6003 | 0.3694 | |
| LIMABackbone=EfficientNetV2-M, Sub-region division=SLICO2025.04 | 0.367 | 0.4575 | 0.4799 | 0.4859 | 0.3518 | |
| LIMABackbone=MobileNetV2, Sub-region division=SLICO2025.04 | 0.3466 | 0.4686 | 0.5462 | 0.6278 | 0.3448 | |
| RISEBackbone=EfficientNetV2-M2025.04 | 0.2958 | 0.3339 | 0.3473 | 0.3697 | 0.2274 | |
| Previous workBackbone=EfficientNetV2-M2025.04 | 0.2616 | 0.3117 | 0.3235 | 0.3306 | 0.1748 | |
| Score-CAM (w/ SMDL-Attribution)base_attribution=Score-CAM, search_algorithm=SMDL-Attribution2024.02 | 0.2491 | 0.3395 | 0.3796 | 0.4082 | 0.1622 | |
| Previous workBackbone=MobileNetV22025.04 | 0.246 | 0.4142 | 0.4913 | 0.5367 | 0.1922 | |
| HSIC-Attribution (w/ SMDL-Attribution)base_attribution=HSIC-Attribution, search_algorithm=SMDL-Attribution2024.02 | 0.243 | 0.3519 | 0.3984 | 0.4513 | 0.1772 | |
| Previous workBackbone=ResNet-1012025.04 | 0.243 | 0.3519 | 0.3984 | 0.4513 | 0.1772 | |
| Grad-CAM++ (w/ SMDL-Attribution)base_attribution=Grad-CAM++, search_algorithm=SMDL-Attribution2024.02 | 0.2424 | 0.3575 | 0.3934 | 0.4193 | 0.1672 | |
| HSIC-AttributionBackbone=EfficientNetV2-M2025.04 | 0.2418 | 0.2561 | 0.2615 | 0.2679 | 0.1611 | |
| RISEBackbone=ResNet-1012025.04 | 0.2406 | 0.3032 | 0.3316 | 0.3807 | 0.2238 | |
| RISEBackbone=MobileNetV22025.04 | 0.234 | 0.3101 | 0.3355 | 0.428 | 0.2573 | |
| Grad-CAM++Backbone=EfficientNetV2-M2025.04 | 0.2338 | 0.2549 | 0.2598 | 0.2659 | 0.1605 | |
| Score-CAMBackbone=EfficientNetV2-M2025.04 | 0.2126 | 0.2327 | 0.2375 | 0.2403 | 0.1572 | |
| SMDL-Attribution (Patch 10x10)grid_size=10x10, priori=None2024.02 | 0.202 | 0.4065 | 0.4908 | 0.5237 | 0.1519 | |
| Grad-CAM++2024.02 | 0.1988 | 0.2447 | 0.2544 | 0.2647 | 0.1094 | |
| Grad-CAM++Backbone=ResNet-1012025.04 | 0.1988 | 0.2447 | 0.2544 | 0.2647 | 0.1094 | |
| Score-CAM2024.02 | 0.1896 | 0.2323 | 0.2449 | 0.251 | 0.1073 | |
| Score-CAMBackbone=ResNet-1012025.04 | 0.1896 | 0.2323 | 0.2449 | 0.251 | 0.1073 | |
| HSIC-Attribution2024.02 | 0.1709 | 0.2091 | 0.225 | 0.2493 | 0.1446 | |
| HSIC-AttributionBackbone=ResNet-1012025.04 | 0.1709 | 0.2091 | 0.225 | 0.2493 | 0.1446 | |
| HSIC-AttributionBackbone=MobileNetV22025.04 | 0.1648 | 0.219 | 0.2415 | 0.2914 | 0.1635 | |
| Grad-CAM++Backbone=MobileNetV22025.04 | 0.1584 | 0.282 | 0.3223 | 0.3462 | 0.1284 | |
| Score-CAMBackbone=MobileNetV22025.04 | 0.1574 | 0.2456 | 0.2948 | 0.3141 | 0.1195 | |
| Kernel SHAPBackbone=MobileNetV22025.04 | 0.0525 | 0.1367 | 0.2321 | 0.4142 | 0.2511 | |
| Kernel SHAPBackbone=EfficientNetV2-M2025.04 | 0.0123 | 0.067 | 0.1489 | 0.3357 | 0.0787 | |
| Kernel SHAPBackbone=ResNet-1012025.04 | 0.0083 | 0.0247 | 0.0574 | 0.2642 | 0.1008 |