Skip to content

Modified version of U-net based on "Convolutional Networks for Biomedical Image Segmentation (Ronneberger et al., 2015)" paper.

License

Notifications You must be signed in to change notification settings

Nikronic/CoarseNet

Repository files navigation

U-Net

u net structure

U-net consists of a contracting path (left side) and an expansive path (right side).

Contracting Phase

The contracting path follows the typical architecture of a convolutional network. It consists of the repeated application of two 3x3 convolutions (unpadded convolutions), each followed by a recti ed linear unit (ReLU) and a 2x2 max pooling operation with stride 2 for downsampling.
At each downsampling step we double the number of feature channels.

Expansive Phase

Every step in the expansive path consists of an upsampling of the feature map followed by a 2x2 convolution ("up-convolution") that halves the number of feature channels, a concatenation with the correspondingly cropped feature map from the contracting path, and two 3x3 convolutions, each followed by a ReLU.
The cropping is necessary due to the loss of border pixels in every convolution. At the final layer a 1x1 convolution is used to map each 64-component feature vector to the desired number of classes. In total the network has 23 convolutional layers.

To allow a seamless tiling of the output segmentation map (see Figure 2), it is important to select the input tile size such that all 2x2 max-pooling operations are applied to a layer with an even x- and y-size.

Network Result Example

example

Implementation

Environment

In this repo, I implemented U-net using Pytorch and Python 3.6.7.

Diff with Paper

Except depth, all parameters are same as paper. Default value of depth in paper is 5 (look at the shape at top of page). So you can have deeper or shallower network by changing deep variable.

Reference

Convolutional Networks for Biomedical Image Segmentation (Ronneberger et al., 2015)

Citation

Please cite this project as Nikan Doosti. (2020). Nikronic/CoarseNet: DOI Release (v0.1-alpha). Zenodo. https://doi.org/10.5281/zenodo.3838582

DOI