vertebra classification files. Dataset for training the deep net and config files
This is the dataset trainning files and trained networks to classify Vertebra in spinal cord.
Using Deep learning to identify and classify Vertebrae into one of the 26 possible classes (C1-S4). The deep net will be able to identify centralized Vertebrae, in a specific size.
In order to train first download mdCNN from: https://github.com/hagaygarty/mdCNN
Then put the 'spine' folder inside mdCNN demo folder.
Unzip the Dataset file (spine.7z) , place 'spine.mat' inside 'spine' folder. This is the main dataset file. It contains images for testing and trainning.
Then call 'demoSpine.m' , the trainning process will begin. You will see the 'successRate=' field strarting from 9.59 % and rising.
====================
In order to classify using the pretrained network first unzip the 'spine.7z' file , then 'Nets.7z'. pleace both files under 'spine' folder.
set matlab folder to 'spine' then run:
addpath('../../Trainning' , '../../mdCNN', '../../utilCode');
spineDataset=load('spine.mat');
load('net.mat');
checkNetwork(net,Inf,spineDataset.images,1);
====================
For 78.0% classification net run:
load('net_maxS.mat');
checkNetwork(net_maxS,Inf,spineDataset.images,1);
wait for several minutes, the results will show