Diffusers
TensorBoard
controlnet
stable-diffusion
medical-imaging
retinal-imaging
diabetic-retinopathy
counterfactual
Instructions to use KylianSu/vessel-bezier-retinal-weights with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use KylianSu/vessel-bezier-retinal-weights with Diffusers:
pip install -U diffusers transformers accelerate
from diffusers import ControlNetModel, StableDiffusionControlNetPipeline controlnet = ControlNetModel.from_pretrained("KylianSu/vessel-bezier-retinal-weights") pipe = StableDiffusionControlNetPipeline.from_pretrained( "fill-in-base-model", controlnet=controlnet ) - Notebooks
- Google Colab
- Kaggle
| Warning: You are sending unauthenticated requests to the HF Hub. Please set a HF_TOKEN to enable higher rate limits and faster downloads. | |
| 设备:cuda:2 | |
| [leak_free] train 剔除 leakage 条数 = 1032 | |
| [leak_free] test 去重剔除条数 = 964,剩余 13237 | |
| [leak_free] val 去重剔除条数 = 958,剩余 13269 | |
| 训练集分布:Normal=55162 DR=59047 共 114209 | |
| 测试集分布:Normal=6896 DR=6341 共 13237 | |
| Epoch 01/15 loss=0.5783 val_auc=0.7917 sens=0.5348 spec=0.9111 f1=0.6558 lr=2.97e-04 | |
| Epoch 02/15 loss=0.5585 val_auc=0.7778 sens=0.5686 spec=0.8619 f1=0.6619 lr=2.87e-04 | |
| Epoch 03/15 loss=0.5373 val_auc=0.7766 sens=0.5446 spec=0.8735 f1=0.6478 lr=2.71e-04 | |
| Epoch 04/15 loss=0.5292 val_auc=0.7859 sens=0.5326 spec=0.9109 f1=0.6540 lr=2.50e-04 | |
| Epoch 05/15 loss=0.5284 val_auc=0.7817 sens=0.5615 spec=0.8712 f1=0.6603 lr=2.25e-04 | |
| Epoch 06/15 loss=0.5204 val_auc=0.7925 sens=0.6009 spec=0.8576 f1=0.6848 lr=1.96e-04 | |
| Epoch 07/15 loss=0.5232 val_auc=0.7972 sens=0.6217 spec=0.8470 f1=0.6958 lr=1.66e-04 | |
| Epoch 08/15 loss=0.5059 val_auc=0.7960 sens=0.6318 spec=0.8290 f1=0.6955 lr=1.34e-04 | |
| Epoch 09/15 loss=0.4882 val_auc=0.8079 sens=0.4978 spec=0.9597 f1=0.6459 lr=1.04e-04 | |
| Epoch 10/15 loss=0.4781 val_auc=0.8159 sens=0.5345 spec=0.9526 f1=0.6741 lr=7.50e-05 | |
| Epoch 11/15 loss=0.4676 val_auc=0.8258 sens=0.5529 spec=0.9516 f1=0.6888 lr=4.96e-05 | |
| Epoch 12/15 loss=0.4579 val_auc=0.8201 sens=0.6046 spec=0.9128 f1=0.7118 lr=2.86e-05 | |
| Epoch 13/15 loss=0.4498 val_auc=0.8352 sens=0.6026 spec=0.9292 f1=0.7177 lr=1.30e-05 | |
| Epoch 14/15 loss=0.4423 val_auc=0.8366 sens=0.6131 spec=0.9192 f1=0.7211 lr=3.28e-06 | |
| Epoch 15/15 loss=0.4373 val_auc=0.8371 sens=0.5894 spec=0.9408 f1=0.7130 lr=0.00e+00 | |
| ============================================================ | |
| 模式:baseline | |
| Test AUC : 0.8295 ← 核心指标 | |
| Test Sensitivity : 0.5553 ← DR 漏诊率 | |
| Test Specificity : 0.9423 | |
| Test F1 : 0.6864 | |
| 混淆矩阵(行=真实,列=预测): | |
| Pred_Normal Pred_DR | |
| Normal 6498 398 | |
| DR 2820 3521 | |
| ============================================================ | |
| 结果已保存至 /root/autodl-tmp/ST/cls_binary_leakfree/baseline_vitb16/ | |