An idea of combining Densenet and Unet
Network structure of Dense-Unet
Comparison of test results of Dense-Unet on medical data set(Isbi 2015 HeLa cells)
method | Vrand | Vinfo |
---|---|---|
U-Net | 0.5676 | 0.4883 |
FCN-8s | 0.5122 | 0.2397 |
E-Net | 0.5940 | 0.5120 |
Dense-Unet | 0.6017 | 0.5735 |
All code was developed and tested on Nvidia RTX2080Ti the following environment.
Python 2.7
opencv3
numpy
tensorflow>=1.0
keras>=1.4(Or newer)
cuda>=8.0
cudnn>=5.0
The documentation is the data set used to test the code
To train our model on the given dataset using:
python unet_xueguan_danboduan_densenet.py