Fully Convolutional Networks for Semantic Segmentation in Keras
- Modify the file train.sh including the paths to multimodal_keras_wrapper library and Keras.
- Prepare the dataset following the same format as sample_data.
- Insert the data paths in config.py and modify any parameter as desired.
- Run ./train.sh to train a model.
Note that the code has been tested using Theano as backend.
The following dependencies are required for using this library:
- Custom Keras fork >= 2.0.9
- Multimodal Keras Wrapper >= 2.1.6
You can donwload a zip file of the source code directly.
Alternatively, you can clone it from GitHub as follows:
git clone https://github.com/beareme/keras_semantic_segmentation.git
For additional information on the Deep Learning library, visit the official web page www.keras.io or the GitHub repository https://github.com/fchollet/keras.
S. Jégou, M. Drozdzal, D. Vazquez, A. Romero, Y. Bengio (2017). The One Hundred Layers Tiramisu: Fully Convolutional Densenets for Semantic Segmentation. IEEE Conference on Computer Vision and Pattern Recognition Workshops, pp. 1175-1183.
O. Ronneberger, P. Fischer, T. Brox (2015). U-net: Convolutional networks for biomedical image segmentation. International Conference on Medical Image Computing and Computer-Assisted Intervention, pp. 234-241.