List your company
ResearchFormerly WizWand

Fine-grained Image Classification on iNaturalist Aves 16-shot 2018 (test)

69.4AccuracyFew-shot FT + POC
46.5252.4658.464.34Dec 10, 2025
Updated 8mo ago

Evaluation results

MethodLinks
Few-shot FT + POCstrategy=finetuning, variant=POC, shots=16, backbone=OpenCLIP ViT-B/32, LMM=Qwen-2.5-VL-7B2025.1269.4
CLAP + POCstrategy=linear probing, variant=POC, shots=16, backbone=OpenCLIP ViT-B/32, LMM=Qwen-2.5-VL-7B2025.1267.8
Few-shot LP + POCstrategy=linear probing, variant=POC, shots=16, backbone=OpenCLIP ViT-B/32, LMM=Qwen-2.5-VL-7B2025.1267.7
CrossModal LP + POCstrategy=linear probing, variant=POC, shots=16, backbone=OpenCLIP ViT-B/32, LMM=Qwen-2.5-VL-7B2025.1266
CLIP-Adapter + POCstrategy=adapter learning, variant=POC, shots=16, backbone=OpenCLIP ViT-B/32, LMM=Qwen-2.5-VL-7B2025.1263.6
FineR + POCstrategy=non-learned, variant=POC, shots=16, backbone=OpenCLIP ViT-B/32, LMM=Qwen-2.5-VL-7B2025.1262.5
CoOp + POCstrategy=prompt learning, variant=POC, shots=16, backbone=OpenCLIP ViT-B/32, LMM=Qwen-2.5-VL-7B2025.1262
MaPLe + POCstrategy=prompt learning, variant=POC, shots=16, backbone=OpenCLIP ViT-B/32, LMM=Qwen-2.5-VL-7B2025.1261
Tip-Adapter + POCstrategy=adapter learning, variant=POC, shots=16, backbone=OpenCLIP ViT-B/32, LMM=Qwen-2.5-VL-7B2025.1259.6
Few-shot FTstrategy=finetuning, variant=expert, shots=16, backbone=OpenCLIP ViT-B/322025.1258.2
CLAPstrategy=linear probing, variant=expert, shots=16, backbone=OpenCLIP ViT-B/322025.1257
Few-shot LPstrategy=linear probing, variant=expert, shots=16, backbone=OpenCLIP ViT-B/322025.1253.8
CrossModal LPstrategy=linear probing, variant=expert, shots=16, backbone=OpenCLIP ViT-B/322025.1253.3
CLIP-Adapterstrategy=adapter learning, variant=expert, shots=16, backbone=OpenCLIP ViT-B/322025.1250.9
CoOpstrategy=prompt learning, variant=expert, shots=16, backbone=OpenCLIP ViT-B/322025.1248.2
Tip-Adapterstrategy=adapter learning, variant=expert, shots=16, backbone=OpenCLIP ViT-B/322025.1247.7
FineRstrategy=non-learned, variant=expert, shots=16, backbone=OpenCLIP ViT-B/322025.1247.5
MaPLestrategy=prompt learning, variant=expert, shots=16, backbone=OpenCLIP ViT-B/322025.1247.4