Image Classification on Agri-Semantics 52k (test)
72.51Overall AccuracyCropVLM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CropVLMmode=Zero-shot classification, protocol=Nearest-neighbour matching in vision-language embedding space, gradient updates=None2026.05 | 72.51 | 72.51 | 29.71 | |
| OpenAI CLIPmode=Zero-shot classification, protocol=Nearest-neighbour matching in vision-language embedding space, gradient updates=None, Backbone=ViT-B/322026.05 | 70.24 | 70.24 | 28.83 | |
| BioCLIP 2mode=Zero-shot classification, protocol=Nearest-neighbour matching in vision-language embedding space, gradient updates=None2026.05 | 67.74 | 67.74 | 31.17 | |
| BioTrove-CLIPmode=Zero-shot classification, protocol=Nearest-neighbour matching in vision-language embedding space, gradient updates=None2026.05 | 51.07 | 51.07 | 36.2 | |
| BioCLIPmode=Zero-shot classification, protocol=Nearest-neighbour matching in vision-language embedding space, gradient updates=None2026.05 | 48.33 | 48.34 | 34.95 | |
| RemoteCLIPmode=Zero-shot classification, protocol=Nearest-neighbour matching in vision-language embedding space, gradient updates=None2026.05 | 42.52 | 42.52 | 27.57 | |
| AgriCLIPmode=Zero-shot classification, protocol=Nearest-neighbour matching in vision-language embedding space, gradient updates=None2026.05 | 4.04 | 4.04 | 14.61 | |
| SigLIP 2mode=Zero-shot classification, protocol=Nearest-neighbour matching in vision-language embedding space, gradient updates=None2026.05 | 3.43 | 3.43 | 16.91 |