A simple implementation of fucos on tensorflow
VGG16 weights, make a folder named "pretrained" and put the file into it.
The datasets for training, validation and testing can be found here.
To train our network from scratch
python fucos.py train path/to/checkpoint/folder
To test an existing model on a folder of images
python fucos.py test path/to/checkpoint/file path/to/testing/image/folder
or only 1 image
python fucos.py test path/to/checkpoint/file path/to/testing/image 0
This implementation doesn't follow exactly what's on the paper. For more information please check out the original paper.