"U-Net: Convolutional Networks for Biomedical Image Segmentation"

Paper: https://arxiv.org/pdf/1505.04597.pdf

There is large consent that successful training of deep networks requires many thousand annotated training samples. In this paper, we present a network and training strategy that relies on the strong use of data augmentation to use the available annotated samples more efficiently. The architecture consists of a contracting path to capture context and a symmetric expanding path that enables precise localization. We show that such a network can be trained end-to-end from very few images and outperforms the prior best method (a sliding-window convolutional network) on the ISBI challenge for segmentation of neuronal structures in electron microscopic stacks. Using the same network trained on transmitted light microscopy images (phase contrast and DIC) we won the ISBI cell tracking challenge 2015 in these categories by a large margin. Moreover, the network is fast. Segmentation of a 512x512 image takes less than a second on a recent GPU. The full implementation (based on Caffe) and the trained networks are available at http://lmb.informatik.uni-freiburg.de/people/ronneber/u-net.

There are 30 satellite pictures of houses and 25 corresponding labels that indicate the roofs. The training set contains of 25 labeled images and the test set is unlabeled. The task is to train a neural network which predicts the roofs of the house based on the training sets.

To solve this task, a U-Net architecture was chosen. The architecture was shown in Figure 1. Since very few training images exist, data augmentation was used. For this, the image was randomly flipped horizontally and vertically and randomly rotated at an angle of one degree. As loss-function tree different loss functions, the dice-loss and pixelwise-softmax and the softmax of tensorflow, are used and compared.

The results are produced under the following conditions:

Hyperparameter: Batch-Size = 8, instance normalization, Cyclic-Learningrate [1e-4,..., 1e-5], epochs = 10000, Data augumentation = flipping [horizontal, vertical] and rotation [0,...,360], Adam with beta1 = 0.9, Filter size of first conv-layer = 64, Unet-depth = 5

1. Clone the repository

$ git clone https://github.com/Shumway82/tf_base.git

2. Go to folder

$ cd tf_base

3. Install with pip3

$ pip3 install -e .

1. Clone the repository

$ https://github.com/Shumway82/U-Net-Segmentation.git

2. Go to folder

$ cd Binary-Classification

3. Install with pip3

$ pip3 install -e .

1. Training

$ python pipeline_trainer.py --dataset "../Data/" --loss "dice"

2. Inferencing

$ python pipeline_inferencer.py --dataset "../Data/" --model_dir "Models_dice"