Medical Image Classification on DermaMNIST
83.6AccuracyMKG-CARE
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MKG-CARECategory=Ours2026.05 | 83.6 | 96.1 | |
| KEMCategory=Knowledge-enhanced, Year=20252026.05 | 81.7 | 95.8 | |
| LatMSFResolution=2242025.07 | 81 | 92.5 | |
| MO-MAE2026.05 | 81 | 95.9 | |
| MSFResolution=282025.07 | 78.8 | 89.6 | |
| MedViT-SResolution=2242025.07 | 78 | 93.7 | |
| MedVIT-S2026.05 | 78 | 93.7 | |
| MedMambaCategory=Vision-based, Year=20242026.05 | 77.9 | 91.7 | |
| AMIMV-SSLBackbone=ResNet-50, Initialization=Random, Training Strategy=Self-supervised2026.04 | 77.85 | — | |
| Swin-TransformerCategory=Vision-based, Year=20212026.05 | 77.8 | 93.8 | |
| MedVIT-L2026.05 | 77.3 | 92 | |
| Google AutoML Vision2026.02 | 76.8 | 91.4 | |
| Google AutoML2026.05 | 76.8 | 91.4 | |
| MedVIT-T2026.05 | 76.8 | 91.4 | |
| Gemini-3-proCategory=General LVLM, Year=20252026.05 | 75.6 | 85.9 | |
| ResNet-18Resolution=2242026.02 | 75.4 | 92 | |
| ResNet-18Resolution=2242025.07 | 75.4 | 92 | |
| ResNet-182026.05 | 75.4 | 92 | |
| EfficientNetV2Category=Vision-based, Year=20212026.05 | 75.1 | 90.1 | |
| AutoKeras2026.02 | 74.9 | 91.5 | |
| AutoKerasResolution=282025.07 | 74.9 | 91.5 | |
| AutoKeras2026.05 | 74.9 | 91.5 | |
| ReSSLBackbone=ResNet-50, Initialization=Random, Training Strategy=Self-supervised2026.04 | 74.75 | — | |
| SimCLRBackbone=ResNet-50, Initialization=Random, Training Strategy=Self-supervised2026.04 | 74.22 | — | |
| ResNet-18Resolution=282026.02 | 73.5 | 91.7 | |
| ResNet-50Resolution=282026.02 | 73.5 | 91.3 | |
| ResNet-18Resolution=282025.07 | 73.5 | 91.7 | |
| ResNet-50Resolution=282025.07 | 73.5 | 91.3 | |
| MoCoV3Backbone=ResNet-50, Initialization=Random, Training Strategy=Self-supervised2026.04 | 73.45 | — | |
| ResNet-50Resolution=2242026.02 | 73.1 | 91.2 | |
| ResNet-50Resolution=2242025.07 | 73.1 | 91.2 | |
| ResNet-502026.05 | 73.1 | 91.2 | |
| VICRegBackbone=ResNet-50, Initialization=Random, Training Strategy=Self-supervised2026.04 | 73.06 | — | |
| Barlow TwinsBackbone=ResNet-50, Initialization=Random, Training Strategy=Self-supervised2026.04 | 72.83 | — | |
| GPT-5.1Category=General LVLM, Year=20252026.05 | 72.7 | 84.3 | |
| NNCLRBackbone=ResNet-50, Initialization=Random, Training Strategy=Self-supervised2026.04 | 72.68 | — | |
| auto-sklearn2026.02 | 71.9 | 90.2 | |
| auto-sklearnResolution=282025.07 | 71.9 | 90.2 | |
| Auto-sklearn2026.05 | 71.9 | 90.2 | |
| G-LoG bi-filtration (MLP)Sigma (σ)=0.52026.02 | 70.8 | 80.6 | |
| G-LoG bi-filtration (MLP)Sigma (σ)=02026.02 | 70.3 | 80.9 | |
| VMambaCategory=Vision-based, Year=20242026.05 | 70.3 | 89 | |
| G-LoG bi-filtration (MLP)Sigma (σ)=12026.02 | 70 | 81.5 | |
| G-LoG bi-filtration (MLP)Sigma (σ)=1.52026.02 | 69.8 | 81.6 | |
| SymmFlowResolution=282025.07 | 69.3 | 83.4 | |
| DINOBackbone=ResNet-50, Initialization=Random, Training Strategy=Self-supervised2026.04 | 68.78 | — | |
| BYOLBackbone=ResNet-50, Initialization=Random, Training Strategy=Self-supervised2026.04 | 68.6 | — | |
| ConvNeXtCategory=Vision-based, Year=20222026.05 | 67 | 91.9 | |
| Topo-Med (MLP)2026.02 | 66.9 | 90.4 | |
| HuatuoGPT-Vision-34BCategory=Med-LVLM, Year=20242026.05 | 61.7 | 65 | |
| Qwen3.5-plusCategory=General LVLM, Year=20262026.05 | 56.4 | 71.9 | |
| medgemma-1.5-4b-itCategory=Med-LVLM, Year=20262026.05 | 54.9 | 60.6 | |
| BioVLMTrain Params=30K2026.04 | 45.27 | — | |
| CoCoOpTrain Params=44K2026.04 | 40.55 | — | |
| TCPTrain Params=0.5M2026.04 | 39.97 | — | |
| PromptSRCTrain Params=69K2026.04 | 39.8 | — | |
| MaPLeTrain Params=5.3M2026.04 | 37.53 | — | |
| BioMedCLIPTrain Params=02026.04 | 36.21 | — | |
| ProGradTrain Params=3K2026.04 | 35.88 | — | |
| Health-GPT-XL32Category=Med-LVLM, Year=20242026.05 | 33.3 | 52.1 | |
| KgCoOpTrain Params=3K2026.04 | 30.09 | — | |
| BiomedCoOpTrain Params=3K2026.04 | 30.02 | — | |
| CoOpTrain Params=3K2026.04 | 28.78 | — | |
| AMIMV-SSLBackbone=ResNet-50, Initialization=random, Training Protocol=self-supervised2026.04 | — | 93.6 | |
| Barlow TwinsBackbone=ResNet-50, Initialization=random, Training Protocol=self-supervised2026.04 | — | 86.28 | |
| BYOLBackbone=ResNet-50, Initialization=random, Training Protocol=self-supervised2026.04 | — | 76.65 | |
| DINOBackbone=ResNet-50, Initialization=random, Training Protocol=self-supervised2026.04 | — | 80.67 | |
| MoCoV3Backbone=ResNet-50, Initialization=random, Training Protocol=self-supervised2026.04 | — | 87.6 | |
| NNCLRBackbone=ResNet-50, Initialization=random, Training Protocol=self-supervised2026.04 | — | 85.44 | |
| ReSSLBackbone=ResNet-50, Initialization=random, Training Protocol=self-supervised2026.04 | — | 89.87 | |
| SimCLRBackbone=ResNet-50, Initialization=random, Training Protocol=self-supervised2026.04 | — | 89.2 | |
| VICRegBackbone=ResNet-50, Initialization=random, Training Protocol=self-supervised2026.04 | — | 86.64 |