Data Dictionary

Augmented Training data

Is a Package where within each hdf5 .h5 file are two groups data/ and labels/ containing the image data and the training labels respectively. T

Augmented Image data

Package

Defined as the set of 16 hdf5 image files each containing the stack of images perturbed 16 different ways.

Example:

$ ls
full_stacks_v10.h5
full_stacks_v11.h5
full_stacks_v12.h5
full_stacks_v13.h5
full_stacks_v14.h5
full_stacks_v15.h5
full_stacks_v16.h5
full_stacks_v1.h5
full_stacks_v2.h5
full_stacks_v3.h5
full_stacks_v4.h5
full_stacks_v5.h5
full_stacks_v6.h5
full_stacks_v7.h5
full_stacks_v8.h5
full_stacks_v9.h5

Predict

Using the Trained Model on real images to segment desired features from images.

Train

The process of teaching the Artificial neural network. For CDeep3M, it involves running caffe with the model in this source tree on training data to learn how to segment desired features from images. This knowledge is stored in a set of files known as a Trained Model

Trained Model

For CDeep3M this is the stored knowledge on how to segment desired features from images. The model comprises (in case of CDeep3M 3 separate models) is stored in .caffemodel files and is created by invocation of runtraining.sh script that is part of CDeep3M.