Closed drinkingkazu closed 7 years ago
I have a motivation to test out the last bullet
*) Single particle classification
en route to training DeconvNet so i'll do that
I will test supera on this new branch, as I have just successfully made images checked in the viewer.
Was planning to test Supera on the GPVM where current data larlite files reside.
Getting this compilation error. Looking into it. But maybe one of you knows where to look.
[ tmw@uboonegpvm03 Base ]$ make
make: *\ No rule to make target install_header', needed by
main_build'. Stop.
in GNUMakefile.CORE
problem comes from this:
install_header: $(OBJECTS) $(DICT).o
@echo '<< installing headers >>'
@$(foreach f, $(HEADERS),$(shell cp $(f) $(PACKAGE_INCDIR)/))
which is a rule that is only defined if ROOT6 exists, but not if ROOT5 exists.
@drinkingkazu says he knows what to do
According to @drinkingkazu this branch really should only be applied to ROOT6. ROOT5 will take some work.
@twongjirad if you can test Supera w/ ROOT6 that would be actually great. I can work on ROOT5 support in the mean time unless you started already.
merged #44
Single particle training seems to work OK. Trained about half an epoch for alex 5 particle high resolution.
OK ROOT5 build is now supported. Checked ROOT6 still builds just fine. Alongside fixed a bug to generate .depend: people has to remove .depend individually unfortunately but then one can benefit from it (avoids re-compilation of some packages under app).
pull request into develop made #55
Actually there will be a backward-incompatible change in caffe configuration file w/ apicaffe_cleanup branch. Might be a good idea to make a tag from current develop.
The branch apicaffe_cleanup has a lot of enhancements (segmentation training and full implementation of ProcessDriver for caffe training) while it is supposed to support old features as well. We have to merge back into master at some point.
We must make sure to test with:
) Supera ) Neutrino + Cosmic overlay image generation ) Semantic Segmentation training ) Single particle classification training