FOCUS-Retinal-Benchmark / data /model_reliability_radar_values.csv
dsrestrepo's picture
Publish static FOCUS benchmark with redesigned research dashboard
be53f07 verified
Raw History Blame Contribute Delete
3.44 kB
family,model_type,model_display,method_display,model_method,accuracy,auc,auprc,ece,age_equalized_odds_gap,image_quality_accuracy_gap,calibration,age_fairness,quality_robustness
cv,cv_general,DINOv2-L,Linear probe,DINOv2-L (linear probing),0.808,0.842,0.627,0.216,0.155,0.109,0.784,0.845,0.891
cv,cv_general,DINOv3-L,Linear probe,DINOv3-L (linear probing),0.809,0.83,0.635,0.213,0.152,0.077,0.787,0.848,0.923
cv,cv_general,ViT-L/16,Linear probe,ViT-L/16 (linear probing),0.791,0.812,0.59,0.198,0.166,0.068,0.802,0.834,0.932
cv,cv_ophthalmo,RETFound CFP,Linear probe,RETFound CFP (linear probing),0.697,0.714,0.431,0.22,0.155,0.18,0.78,0.845,0.82
cv,cv_ophthalmo,RETFound DINOv2 MEH,Linear probe,RETFound DINOv2 MEH (linear probing),0.76,0.5,0.24,0.291,0.0,0.049,0.709,1.0,0.951
cv,cv_ophthalmo,RETFound DINOv2 SH,Linear probe,RETFound DINOv2 SH (linear probing),0.76,0.5,0.24,0.291,0.0,0.049,0.709,1.0,0.951
cv,cv_ophthalmo,RETFound MAE MEH,Linear probe,RETFound MAE MEH (linear probing),0.675,0.688,0.423,0.234,0.192,0.165,0.766,0.808,0.835
cv,cv_ophthalmo,RETFound MAE SH,Linear probe,RETFound MAE SH (linear probing),0.695,0.704,0.44,0.223,0.183,0.128,0.777,0.817,0.872
cv,cv_ophthalmo,RETFound OCT,Linear probe,RETFound OCT (linear probing),0.694,0.698,0.412,0.203,0.121,0.065,0.797,0.879,0.935
cv,cv_ophthalmo,VisionFM,Linear probe,VisionFM (linear probing),0.849,0.862,0.711,0.153,0.156,0.046,0.847,0.844,0.954
mllm,mllm_general,Gemma-3-27B,Base,Gemma-3-27B (base),0.427,0.648,0.392,0.568,0.086,0.033,0.432,0.914,0.967
mllm,mllm_general,LLaVA-NeXT-8B,Base,LLaVA-NeXT-8B (base),0.703,0.593,0.352,0.238,0.024,0.033,0.762,0.976,0.967
mllm,mllm_general,Qwen3-VL-8B,Base,Qwen3-VL-8B (base),0.594,0.733,0.527,0.359,0.076,0.006,0.641,0.924,0.994
mllm,mllm_medical,MedGemma 1.5 4B,Base,MedGemma 1.5 4B (base),0.79,0.816,0.668,0.181,0.086,0.018,0.819,0.914,0.982
mllm,mllm_medical,MedGemma 27B,Base,MedGemma 27B (base),0.803,0.777,0.628,0.187,0.061,0.018,0.813,0.939,0.982
mllm,mllm_medical,MedGemma 4B,Base,MedGemma 4B (base),0.781,0.818,0.629,0.215,0.132,0.033,0.785,0.868,0.967
vlm,vlm_general,CLIP-B/32,Linear probe,CLIP-B/32 (linear probing),0.767,0.795,0.544,0.235,0.125,0.21,0.765,0.875,0.79
vlm,vlm_general,CLIP-B/32,Zero-shot,CLIP-B/32 (zero shot),0.379,0.516,0.28,0.336,0.096,0.11,0.664,0.904,0.89
vlm,vlm_general,MedSigLIP,Linear probe,MedSigLIP (linear probing),0.874,0.908,0.757,0.126,0.082,0.052,0.874,0.918,0.948
vlm,vlm_general,MedSigLIP,Zero-shot,MedSigLIP (zero shot),0.672,0.583,0.339,0.205,0.064,0.055,0.795,0.936,0.945
vlm,vlm_general,SigLIP2-B,Linear probe,SigLIP2-B (linear probing),0.764,0.784,0.54,0.242,0.117,0.201,0.758,0.883,0.799
vlm,vlm_general,SigLIP2-B,Zero-shot,SigLIP2-B (zero shot),0.605,0.549,0.289,0.283,0.0,0.049,0.717,1.0,0.951
vlm,vlm_ophthalmo,EyeCLIP,Linear probe,EyeCLIP (linear probing),0.565,0.57,0.286,0.285,0.218,0.098,0.715,0.782,0.902
vlm,vlm_ophthalmo,EyeCLIP,Zero-shot,EyeCLIP (zero shot),0.547,0.516,0.259,0.25,0.0,0.049,0.75,1.0,0.951
vlm,vlm_ophthalmo,FLAIR,Linear probe,FLAIR (linear probing),0.849,0.872,0.726,0.134,0.107,0.035,0.866,0.893,0.965
vlm,vlm_ophthalmo,FLAIR,Zero-shot,FLAIR (zero shot),0.74,0.732,0.482,0.237,0.051,0.051,0.763,0.949,0.949
vlm,vlm_ophthalmo,RET-CLIP,Linear probe,RET-CLIP (linear probing),0.875,0.877,0.763,0.126,0.099,0.057,0.874,0.901,0.943
vlm,vlm_ophthalmo,RET-CLIP,Zero-shot,RET-CLIP (zero shot),0.622,0.693,0.444,0.234,0.094,0.076,0.766,0.906,0.924