try:
args.append(pkl.load(f))
except EOFError:
break
scaling = global_params.config.entries['Dataset']['scaling']
for cc in args:
svixs = list(cc.nodes())
cc_ix = np.min(svixs)
sso = SuperSegmentationObject(cc_ix, version="gliaremoval", nb_cpus=2,
working_dir=global_params.config.working_dir,
create=True, scaling=scaling,
sv_ids=svixs)
so_cc = nx.Graph()
for e in cc.edges():
so_cc.add_edge(sso.get_seg_obj("sv", e[0]),
sso.get_seg_obj("sv", e[1]))
sso._rag = so_cc
sd = sos_dict_fact(svixs)
sos = init_sos(sd)
sso._objects["sv"] = sos
try:
sso.gliasplit(verbose=False)
except Exception as e:
print("\n-------------------------------------\n"
"Splitting of SSV %d failed with %s."
"\n-------------------------------------\n" % (cc_ix, e))
with open(path_out_file, "wb") as f:
pkl.dump("0", f)
args.append(pkl.load(f))
except EOFError:
break
scaling = global_params.config['scaling']
# TODO: This coulb be cunked by loading `mesh_bb` and glia prob. prediction cache arrays
# (might have to be create via `dataset_analysis`)
for cc in args:
svixs = list(cc.nodes())
cc_ix = np.min(svixs)
sso = SuperSegmentationObject(cc_ix,
version="gliaremoval",
nb_cpus=1,
working_dir=global_params.config.working_dir,
create=True,
scaling=scaling,
sv_ids=svixs)
so_cc = nx.Graph()
for e in cc.edges():
so_cc.add_edge(sso.get_seg_obj("sv", e[0]),
sso.get_seg_obj("sv", e[1]))
sso._rag = so_cc
sd = sos_dict_fact(svixs)
sos = init_sos(sd)
sso._objects["sv"] = sos
sso.load_attr_dict()
sso.gliasplit(verbose=False, recompute=False)
with open(path_out_file, "wb") as f:
pkl.dump("0", f)