Text-based Person Anomaly Search on PAB 1.0 (test)
87.21R@1 ScoreSSDC
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| SSDCImage Enc.=ViT-B/16, Text Enc.=BERT-Base, Protocol=Fine-tuned with 1M PAB2026.04 | 87.21 | 99.09 | 99.75 | 92.87 | |
| CMPImage Enc.=ViT-B/16, Text Enc.=BERT-Base, Protocol=Fine-tuned with 1M PAB2026.04 | 84.93 | 99.09 | 99.75 | 91.66 | |
| RDE + Detective Squad InteractionImage Enc.=ViT-B/16, Text Enc.=Transformer, Protocol=Fine-tuned with 1M PAB2026.04 | 84.88 | 98.13 | 98.69 | 91.08 | |
| IRRA + Detective Squad InteractionImage Enc.=ViT-B/16, Text Enc.=Transformer, Protocol=Fine-tuned with 1M PAB2026.04 | 83.92 | 98.33 | 99.09 | 90.6 | |
| CMPImage Enc.=ViT-B/16, Text Enc.=BERT-Base, Protocol=Fine-tuned with 0.1M PAB2026.04 | 83.06 | 98.89 | 99.49 | 90.41 | |
| X-VLMImage Enc.=Swin-B, Text Enc.=BERT-Base, Protocol=Fine-tuned with 0.1M PAB2026.04 | 81.95 | 98.84 | 99.19 | 89.86 | |
| RaSaImage Enc.=ViT-B/16, Text Enc.=BERT-Base, Protocol=Fine-tuned with 0.1M PAB2026.04 | 80.79 | 98.89 | 99.65 | 89.2 | |
| IRRAImage Enc.=ViT-B/16, Text Enc.=Transformer, Protocol=Fine-tuned with 1M PAB2026.04 | 78.67 | 97.98 | 98.94 | 87.74 | |
| CLIPImage Enc.=ViT-B/16, Text Enc.=Transformer, Protocol=Fine-tuned with 0.1M PAB2026.04 | 77.6 | 98.84 | 99.75 | 87.35 | |
| RDEImage Enc.=ViT-B/16, Text Enc.=Transformer, Protocol=Fine-tuned with 1M PAB2026.04 | 76.74 | 96.97 | 98.38 | 86.12 | |
| IRRAImage Enc.=ViT-B/16, Text Enc.=Transformer, Protocol=Fine-tuned with 0.1M PAB2026.04 | 76.39 | 97.62 | 99.14 | 86.33 | |
| WoRAImage Enc.=Swin-B, Text Enc.=BERT-Base, Protocol=Fine-tuned with 0.1M PAB2026.04 | 74.47 | 96.82 | 98.48 | 84.6 | |
| CAMeLImage Enc.=SG-Former, Text Enc.=BERT-Base, Protocol=Fine-tuned with 0.1M PAB2026.04 | 74.3 | 96.79 | 98.84 | 84.2 | |
| APTMImage Enc.=Swin-B, Text Enc.=BERT-Base, Protocol=Fine-tuned with 0.1M PAB2026.04 | 72.14 | 95.3 | 97.17 | 82.78 | |
| X-VLMImage Enc.=Swin-B, Text Enc.=BERT-Base, Protocol=Zero-shot Performance2026.04 | 71.94 | 97.78 | 98.99 | 83.96 | |
| MRAImage Enc.=Swin-B, Text Enc.=BERT-Base, Protocol=Fine-tuned with 0.1M PAB2026.04 | 70.53 | 94.69 | 97.47 | 81.59 | |
| RDE + Detective Squad InteractionImage Enc.=ViT-B/16, Text Enc.=Transformer, Protocol=Zero-shot Performance2026.04 | 63.55 | 78.56 | 81.5 | 71 | |
| IRRA + Detective Squad InteractionImage Enc.=ViT-B/16, Text Enc.=Transformer, Protocol=Zero-shot Performance2026.04 | 60.57 | 75.08 | 81.45 | 68 | |
| CLIPImage Enc.=ViT-B/16, Text Enc.=Transformer, Protocol=Zero-shot Performance2026.04 | 47.57 | 81.55 | 89.03 | 62.73 | |
| RDEImage Enc.=ViT-B/16, Text Enc.=Transformer, Protocol=Zero-shot Performance2026.04 | 41.3 | 69.01 | 77.35 | 54.36 | |
| IRRAImage Enc.=ViT-B/16, Text Enc.=Transformer, Protocol=Zero-shot Performance2026.04 | 30.59 | 59.61 | 68.91 | 44.41 | |
| CAMeLImage Enc.=SG-Former, Text Enc.=BERT-Base, Protocol=Zero-shot Performance2026.04 | 24.47 | 50 | 58.75 | 36.75 | |
| APTMImage Enc.=Swin-B, Text Enc.=BERT-Base, Protocol=Zero-shot Performance2026.04 | 22.9 | 45.8 | 52.38 | 33.56 | |
| WoRAImage Enc.=Swin-B, Text Enc.=BERT-Base, Protocol=Zero-shot Performance2026.04 | 22.25 | 45.91 | 53.54 | 33.39 | |
| RaSaImage Enc.=ViT-B/16, Text Enc.=BERT-Base, Protocol=Zero-shot Performance2026.04 | 21.74 | 27.3 | 27.96 | 24.35 | |
| MRAImage Enc.=Swin-B, Text Enc.=BERT-Base, Protocol=Zero-shot Performance2026.04 | 9.91 | 23.66 | 31.45 | 17.15 |