First time configuration:
git clone https://github.com/NevisUB/ssnet_example
cd ssnet_example/sw
source setup.sh
Above "source setup.sh" pulls repositories and builds. caffe and larcv has a particular order (and repeats) of builds, so please follow this procedure of build. You can look at setup.sh if you are curious of this "order and repeats": it's ultimately something I should fix...
From 2nd time (like you logged out and logged in again), just do
source setup.sh
under the "sw" directory. This won't attempt a build and you will be ready to use a software.
You need a tagger input (holds original image data) and output (holds tagger's data).
python run_ssnet.py out.root FILE1 [FILE2 ...]
where FILE1, FILE2, etc. are all input files.
out.root is going to be the name of the output file.
If the same named file already exsits it halt the execution and exits.
First, set up
(log in)
> cd WHERE_YOU_PUT_THIS_REPO
> source sw/setup.sh
We will use one version of trained network. The script we will use (run_ssnet.py) requires trained networks' weights to exist in the same directory. So let's make symbolic links there.
ln -s /data/drinkingkazu/UBDeconvNet/dlmc_mcc8_ssnet_v4/segmentation_pixelwise_ikey_plane0_iter_75500.caffemodel
ln -s /data/drinkingkazu/UBDeconvNet/dlmc_mcc8_ssnet_v4/segmentation_pixelwise_ikey_plane1_iter_65500.caffemodel
ln -s /data/drinkingkazu/UBDeconvNet/dlmc_mcc8_ssnet_v4/segmentation_pixelwise_ikey_plane2_iter_68000.caffemodel
... where three symbolic links correspond to a stored weight file per plane (you see plane0, plane1, plane2 in file name).
Next let's make a sample input root file with just 5 events. We just use one of many handy example files @ wu to do this.
run_processor copy.cfg /stage2/drinkingkazu/march23/intrinsic_nue/out_supera/larcv_0000_0099.root
Now we are ready to try out
python run_ssnet.py out.root sample.root
... where out.root is the output file name and sample.root is the input (symbolic link) file.