- Log onto Della and navigate to the directory where yo want to keep the scripts
- Git clone this repo to your folder by
git clone https://github.com/cyhFlight/wang_mouseBehavior
Alternatively, just copy everything from my scripts folder by
cp /tigress/yuhang/wang_lab/scripts /YOURFOLDER
- Copy folder wang_mouseBehavior/bin/ to your own HOME folder
cp codes/wang_mouseBehavior/bin/ ~/
cd /wang_mouseBehavior/sbatch_scripts/sbatchTemplates/
Edit all .sbTemp files. Change the #SBATCH line about the email to receive notifications.
- On Della, navigate to folder with all movies to process, make symbolic links for sbatch tamplates by
ln -s ~/codes/wang_mouseBehavior/sbatch_scripts/sbatchTemplates/ ./
- Run (by just entering)
init_dir.sh
This will generate all the List.txt files and sbatch files you need. 3. On DELLA submit the job via slurm
sbatch 1_runBox.sb
- Wait until the job finishes. You should get emails about it. If fails, check the slurm output files under /projects/WANG/mouseBehavior/slurm_logs/ for error messages
*After it is done, check box movies in savePath by plotting mouseDataBox from '*_box_data.mat' *check videos with plotcentroidbox.JV. You need to open matlab as:
/usr/licensed/bin/matlab-R2017b -nosplash -nodesktop -nodisplay -singleCompThread -r MATLAB_command
- Switch to TIGER_GPU, and go to the same folder, submit the next job
sbatch 2_predAlignMouse.sb
- Wait for the job to finish. Then switch back to Della folder to submit the next job
sbatch 3_runMouseAlign.sb
- Wait for the job to finish. *check videos with plotcentroidaligned.JV with MATLAB
/usr/licensed/bin/matlab-R2017b -nosplash -nodesktop -nodisplay -singleCompThread -r MATLAB_command
- Log onto TIGER_GPU, in the same folder to submit
sbatch 4_predMultMouse_JV.sb
- The output .h5 files are all in leapout/ folder.
LEAP network for alignments: 181023_131517-n=399
LEAP network for labeling joints: mouseBox_trainingSet_cluster_6-leap_cnn_epochs=50
Then locally open the masterscript_DREADD
-
Run FromFilesIDstoCellaray.m
-
run masterScript_DREADD.m This part may take a long time.
-
Make text file using cmd: navigate to folder by just typing the folder whole path then: dir *.mat /s /b > printit.txt 4)run Distance -> PCA Variance explained to select number of projections (missing code) 5)PCA projections: plot(1-(vals./sum(vals))) 6)Go to SPOCK to run on cluster (spock.princeton.edu) edit runSubs_jv.sh if needed only 20 projections printf '%s\n' "$PWD"/*20.mat > projections20.txt sbatch runSubs_jv.sh (squeue -u jverpeut to check progress)
-
Make the trainingsetData list which is next in the code locally
-
K means clusting: In cluster utilities...edit runC_JV.sh run in spock navigate to folder: cd /jukebox/wang_mkislin/OldData/Jess/code_and_dataset/mouse-behavior/cluster_utilities sbatch runC_jv.sh 1:10, MaxIter 100, replicate 20, 54 hours
-
open runRE_HDK_jv.sh for editing change SBATCH array for size change all paths
-
put all aligned mat files into one folder