Binary Classification on Binary Scar Image Dataset (5-fold cross-validation)
73AccuracyScaFE (RF)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| ScaFE (RF)Features=LLM-guided feature engineering (ScaFE), Downstream Classifier=Random Forest2026.06 | 73 | 72 | 72 | 72 | |
| ScaFE (DT)Features=LLM-guided feature engineering (ScaFE), Downstream Classifier=Decision Tree2026.06 | 69 | 69 | 69 | 68 | |
| ScaFE (SVM)Features=LLM-guided feature engineering (ScaFE), Downstream Classifier=SVM (RBF kernel)2026.06 | 68 | 68 | 68 | 67 | |
| Handcrafted+RFFeatures=Classical (color histograms, GLCM texture, shape descriptors), Downstream Classifier=Random Forest2026.06 | 66 | 65 | 65 | 64 | |
| CNN-ResNet18Backbone=ResNet-18, Pre-training=ImageNet, Status=fine-tuned2026.06 | 61 | 62 | 62 | 55 | |
| MLLM-DirectModel=GPT-4V, Setting=zero-shot2026.06 | 60 | 60 | 60 | 56 | |
| CNN-EfficientNetBackbone=EfficientNet-B0, Pre-training=ImageNet, Status=fine-tuned2026.06 | 58 | 58 | 58 | 55 | |
| ViT-BaseBackbone=Vision Transformer (ViT-Base), Pre-training=trained from scratch2026.06 | 51 | 50 | 50 | 34 |