chm_singularity

Generates Singularity image for Cascaded Hierarchical Model (CHM) version 2.1.367. To make image creation easier a Vagrant configuration is provided that spins up a Virtualbox virtual machine with Singularity installed allowing the steps to be done on any machine with Vagrant and Virtualbox

Run requirements

• Singularity 2.2+

Build requirements

• Vagrant
• VirtualBox
• Git To clone repo, otherwise just download and unzip the repo
• At least 4+gb ram
• At least 20gb free disk space

To Build

After installing Vagrant and Virtualbox, run the following commands to get a Virtual Machine with Singularity installed and connect to it via ssh

# clone repo or download repo
git clone https://github.com/CRBS/chm_singularity.git
cd chm_singularity
cd vagrant
vagrant up
vagrant ssh

Now checkout the repo in the Virtual Machine and build. The image will be created under build/ directory and the commands below copy that to the shared directory /vagrant which is the directory where vagrant up command was issued.

git clone https://github.com/CRBS/chm_singularity.git
cd chm_singularity
make singularity22
mv build/chm_s22.img /vagrant/.
exit

# to get rid of virtual machine run this command:
vagrant destroy 

To test

The image is built with a self test mode. To run the self test issue the following command after building:

# Assuming one is still in Virtual Machine
cd /vagrant
./chm_s22.img verify `pwd`

Expected output from above command

Extracting features ... stage 1 level 0
Start learning LDNN ... stage 1 level 0
Run clustering...Done. It took 0.280864
Number of training samples = 19000
Epoch No. 1 ... error = 0.133863
Epoch No. 2 ... error = 0.109276
Epoch No. 3 ... error = 0.100021
Epoch No. 4 ... error = 0.094533
Epoch No. 5 ... error = 0.090461
Epoch No. 6 ... error = 0.086922
Epoch No. 7 ... error = 0.083974
Epoch No. 8 ... error = 0.081532
Epoch No. 9 ... error = 0.079116
Epoch No. 10 ... error = 0.076921
Epoch No. 11 ... error = 0.075118
Epoch No. 12 ... error = 0.073202
Epoch No. 13 ... error = 0.071481
Epoch No. 14 ... error = 0.070160
Epoch No. 15 ... error = 0.068497
Generating outputs ... stage 1 level 0
Extracting features ... stage 1 level 1
Start learning LDNN ... stage 1 level 1
Run clustering...Done. It took 0.029079
Number of training samples = 4800
Epoch No. 1 ... error = 0.147385
Epoch No. 2 ... error = 0.132599
Epoch No. 3 ... error = 0.129089
Epoch No. 4 ... error = 0.128801
Epoch No. 5 ... error = 0.123766
Epoch No. 6 ... error = 0.124758
Epoch No. 7 ... error = 0.123334
Epoch No. 8 ... error = 0.118936
Epoch No. 9 ... error = 0.122070
Epoch No. 10 ... error = 0.120665
Epoch No. 11 ... error = 0.116100
Epoch No. 12 ... error = 0.119207
Epoch No. 13 ... error = 0.114560
Epoch No. 14 ... error = 0.116833
Epoch No. 15 ... error = 0.117144
Generating outputs ... stage 1 level 1
Extracting features ... stage 2 level 0
Start learning LDNN ... stage 2 level 0
Run clustering...Done. It took 0.411123
Number of training samples = 19000
Epoch No. 1 ... error = 0.109903
Epoch No. 2 ... error = 0.107199
Epoch No. 3 ... error = 0.106083
Epoch No. 4 ... error = 0.105375
Epoch No. 5 ... error = 0.104883
Epoch No. 6 ... error = 0.104358
Generating outputs ... stage 2 level 0
Verifying probability map exists with size greater then 0

To use

To train CHM

The example below uses test data to run chm train putting the results in the model directory of the current working directory.

./chm_s22.img train ../testdata/images ../testdata/labels -S 2 -L 1 -m ./model

To run CHM aka test

The example below uses the trained model from the previous step to run CHM on an image in testdata directory. The resulting probability map is stored in results directory under current working directory.

./chm_s22.img test ../testdata ./result -b 100x95 -t 1,1 -o 0x0 -h -m ./model

LICENSE

For the code to generate to generate this singularity image see the license in this file: LICENSE.txt

For CHM license please visit: https://www.sci.utah.edu/software/chm.html

The scripts in this source tree download and include the Matlab Compiled Runtime 2013a. The license can be found here or by running the following command on the built image:

./chm_s22.img license