For biomedical images, unet was first mentioned in article U-Net: Convolutional Networks for Biomedical Image Segmentation (Ronneberger et al., 2015). In this repository, neuronal membrane segmentation will be perform use the unet architecture in below.
If you want more information about unet, you should to examine this: Freiburg University - Vision
Border reflect is applied to the input images (512x512) to convert the images to 572 x 572 dimensions. Target images were cropped from the center at 388 x 388.
| 512x512 - train/0.tiff | 572x572 border reflect | 388x388 ground truth |
|---|---|---|
 |
 |
 |
When test/0.tiff is given to 300 Epoch trained model inputs, the following result is obtained.
| 512x512 - test/0.tiff | 388x388 model output |
|---|---|
 |
 |