Street-view Image Attribute Classification on GSS Platform
69.12AccuracyCLIP-MHAdapter
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| CLIP-MHAdapterParadigm=Parameter-Efficient Adaptation, Trainable Parameters=1.38M2026.02 | 69.12 | 60.79 | 67.27 | 64.93 | |
| MaxViTParadigm=Vision Transformer, Trainable Parameters=30.9M2026.02 | 68.28 | 56.69 | 69.21 | 49.87 | |
| CLIP-AdapterParadigm=Parameter-Efficient Adaptation, Trainable Parameters=0.52M2026.02 | 68.12 | 57.15 | 69.21 | 71.44 | |
| CoOpParadigm=Parameter-Efficient Adaptation, Trainable Parameters=8K2026.02 | 65.04 | 58.82 | 61.64 | 65.82 | |
| CLIP-Linear ProbeParadigm=Parameter-Efficient Adaptation, Trainable Parameters=3K2026.02 | 63.14 | 52.88 | 64.2 | 66.11 | |
| Zero-shot CLIPParadigm=Zero-shot Transfer2026.02 | 60.98 | 43.19 | 60.8 | 45.99 | |
| ZeroR-TrainerParadigm=Zero-shot Transfer2026.02 | 31.69 | 8.02 | 15.25 | 0 |