keras-unet-collection 0.0.4

The tensorflow.keras implementations of U-net, U-net++, Residual U-net, Attention U-net.

keras-unet-collection

This repository contains tensorflow.keras implementations of U-net, U-net++, R2U-net, Attention U-net, ResUnet-a:

keras_unet_collection.models	Reference
U-net/Unet	Ronneberger et al. (2015)
U-net++/Unet++	Zhou et al. (2018)
R2U-Net	Alom et al. (2018)
Attention U-net	Oktay et al. (2018)
ResUnet-a	Diakogiannis et al. (2020)

These models are implemented with user friendly key words, including optional network depth, hidden layer activations and batch normalization. Examples refers to the user guide.

Additional activation layers and loss functions are also provided:

keras_unet_collection.activations	Reference
Gaussian Error Linear Units (GELU)	Hendrycks et al. (2016)
Snake activation	Liu et al. (2020)

keras_unet_collection.losses	Reference
Tversky loss	Hashemi et al. (2018)
Focal Tversky loss	Abraham et al. (2019)
CRPS loss (experimental)

Dependencies

• TensorFlow 2.3.0

• Keras 2.4.0

• Numpy 1.18.2

Installation and usage

pip install keras-unet-collection

from keras_unet_collection import models
# e.g. models.unet_2d(...)

• Jupyter notebooks are provided as user guides.

• Changlog

Overview

U-net is a convolutional neural network with encoder-decoder architecture and skip-connections, loosely defined under the concept of "fully convolutional networks." U-net was originally proposed for the semantic segmentation of medical images and is modified for solving a wider range of gridded learning problems.

U-net and many of its variants take three or four-dimensional tensors as inputs and produce outputs of the same shape. One technical highlight of these models is the skip-connections from downsampling to upsampling layers that benefit the reconstruction of high-resolution, gridded outputs.

Contact

Yingkai (Kyle) Sha [e-mail]

License

MIT License