Table 10 Unseen ArtDL classification results
From: Multimodal prototype fusion network for paper-cut image classification
Methods | Type | Training | n-shot accuracy (%) | |
|---|---|---|---|---|
5-shot | 16-shot | |||
COOP | Prompt learning | Yes | 25.60 | 26.30 |
CoCoOp | Prompt learning | Yes | 30.20 | 35.10 |
CLIP Zero-shot | Contrastive learning | No | 44.51 | 44.51 |
CLIP-Linear | Deep learning | Yes | 20.13 | 27.70 |
Prototypical Network | Metric learning | No | 40.13 | 47.70 |
Matching Network | Metric learning | No | 30.14 | 32.11 |
AMPFN | Metric learning | Yes | 58.59 | 62.75 |
IMPFN | Metric learning | No | 46.58 | 56.70 |
