Anomaly Segmentation on KSDD2
76.6APAG-VAS
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| AG-VASBase Model=LLaVA-OneVision-7B2026.03 | 76.6 | — | — | — | — | 74.7 | 53.8 | |
| VCP-CLIP (ours)Training Requirement=Training-required2024.07 | 75.2 | 99.5 | 98 | — | — | — | — | |
| Bayes-PFL*Base Model=CLIP(ViT-L-14-336), Retrained=true2026.03 | 74.9 | — | — | — | — | 67.6 | 39.5 | |
| Bayes-PFLBase Model=CLIP(ViT-L-14-336)2026.03 | 73.7 | — | — | — | — | 63.5 | 25.4 | |
| Baseline (ours)Training Requirement=Training-required2024.07 | 71.6 | 99.4 | 95.4 | — | — | — | — | |
| CoCoOpTraining Requirement=Training-required2024.07 | 69.6 | 96.1 | 90.9 | — | — | — | — | |
| LISA*Base Model=LLaVA-OneVision-7B, Retrained=true2026.03 | 69.4 | — | — | — | — | 69.8 | 48.4 | |
| APRIL-GANTraining Requirement=Training-required2024.07 | 67.2 | 97.5 | 49.6 | — | — | — | — | |
| AG-VASBase Model=LLaVA-13B2026.03 | 66.5 | — | — | — | — | 65.4 | 53.7 | |
| LISA*Base Model=LLaVA-13B, Retrained=true2026.03 | 63.7 | — | — | — | — | 63.8 | 50.9 | |
| APRIL-GANBase Model=CLIP(ViT-L-14-336)2026.03 | 61.6 | — | — | — | — | 62.6 | 41.8 | |
| LISABase Model=LLaVA-OneVision-7B2026.03 | 59.8 | — | — | — | — | 65.8 | 49.4 | |
| AnomalyCLIP*Base Model=CLIP(ViT-L-14-336), Retrained=true2026.03 | 48.8 | — | — | — | — | 53.6 | 28.6 | |
| PixelLMBase Model=LLaVA-13B2026.03 | 42.6 | — | — | — | — | 46.8 | 32.4 | |
| AnomalyCLIPBase Model=CLIP(ViT-L-14-336)2026.03 | 41.8 | — | — | — | — | 50.3 | 27.6 | |
| LISABase Model=LLaVA-13B2026.03 | 36.8 | — | — | — | — | 39.8 | 29.4 | |
| AnVoLTraining Requirement=Training-free2024.07 | 33.9 | 95.9 | 80.4 | — | — | — | — | |
| AnomalyGPTTraining Requirement=Training-required2024.07 | 29.7 | 91.5 | 61.9 | — | — | — | — | |
| LISABase Model=LLaVA-7B2026.03 | 19.5 | — | — | — | — | 23.8 | 18.2 | |
| WinCLIPTraining Requirement=Training-free2024.07 | 17.5 | 89.4 | 65.9 | — | — | — | — | |
| WinCLIPBase Model=CLIP(ViT-B-16-240)2026.03 | 17.1 | — | — | — | — | 24.6 | 0 | |
| PaDTBase Model=Qwen2.5-VL-7B2026.03 | 1.2 | — | — | — | — | 2.9 | 12.4 | |
| AdaCLIPTraining setting=Training-based2026.01 | — | 85.8 | 72.9 | — | — | — | — | |
| AnomalyCLIPTraining setting=Training-based2026.01 | — | 97.9 | 94.9 | — | — | — | — | |
| ClipSegZero-shot=true2023.05 | — | — | — | 34.27 | 9.05 | — | — | |
| FAPromptTraining setting=Training-based2026.01 | — | 97.4 | 93.2 | — | — | — | — | |
| MRAD-CLIPTraining setting=Training-based2026.01 | — | 98.9 | 95.6 | — | — | — | — | |
| MRAD-FTTraining setting=Training-based2026.01 | — | 98.8 | 89.8 | — | — | — | — | |
| MRAD-TFTraining setting=Train-free2026.01 | — | 94.9 | 87.1 | — | — | — | — | |
| SAAZero-shot=true2023.05 | — | — | — | 8.79 | 16.4 | — | — | |
| SAA+Zero-shot=true2023.05 | — | — | — | 59.19 | 39.34 | — | — | |
| UTADZero-shot=true2023.05 | — | — | — | 22.53 | 3.56 | — | — | |
| WinCLIPTraining setting=Train-free2026.01 | — | 75.4 | 69.2 | — | — | — | — |