UNet++ on Cell Nuclei Dataset (DSB 2018)
Results (5-fold Cross Validation)
| 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%
Usage
from tensorflow.keras.models import load_model
import helper_functions
model = load_model('best_model_fold1.h5', custom_objects={
'bce_dice_loss': helper_functions.bce_dice_loss,
'dice_coef': helper_functions.dice_coef
})
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support