FCN-pytorch

This is an implementation of Fully Convolutional Networks using pytorch.(https://openaccess.thecvf.com/content_cvpr_2015/html/Long_Fully_Convolutional_Networks_2015_CVPR_paper.html)

In the following, there is a predicted mask with image and groundtruth (In order from left to right: Eye image-Groundtruth image-Predicted mask):

1. requirements

pip install -r requirements.txt

2. images and masks

Put images and maks in dataset/images and dataset/masks

3. train.txt and val.txt

put images name in here:

image_name_1

image_name_2

.

.

.

image_nameN

4. image size

The default image size is 256X256

I will be happy to help you. Please feel free to contact me.