Yuecong682/unetpp-results

UNet++ Reproduction Results

5-Fold Cross Validation Results (Cell Nuclei Dataset)

Fold	Dice	IoU
1	0.8906	0.8451
2	0.9071	0.8480
3	0.9049	0.8519
4	0.8984	0.8478
5	0.9030	0.8441

Mean IoU: 84.74% Mean Dice: 90.08%

Files

• training_curves.png - Training curves visualization
• *.pkl - Detailed results for each branch (L1-L4)
• best_model_fold*.h5 - Model weights (5-fold)

Paper Reference

UNet++: A Nested U-Net Architecture for Medical Image Segmentation Zongwei Zhou, Md Mahfuzur Rahman Siddiquee, Nima Tajbakhsh, Jianming Liang DLMIA 2018