def generate_outfile_names(vid, seglen, step, nframes, nreps, stepint, cutoff):
outroot = os.path.dirname(
mousezopt.generate_outfile_names(vid, seglen, 0, nframes, None, None,
None)[1])
scorefile = os.path.join(
outroot,
'seg%s_nframes%s_step%s_nreps%s_stepint%s_cutoff%s-score.dict' %
(seglen, nframes, step, nreps, stepint, cutoff))
distsfile = os.path.join(
outroot,
'seg%s_nframes%s_step%s_nreps%s_stepint%s_cutoff%s-model_dists.dict' %
(seglen, nframes, step, nreps, stepint, cutoff))
return scorefile, distsfile
def generate_outfile_names(vid,seglen,step,nframes,nreps,stepint,cutoff):
outroot = os.path.dirname(mousezopt.generate_outfile_names(vid,seglen,0,nframes,None,None,None)[1])
scorefile = os.path.join(outroot,'seg%s_nframes%s_step%s_nreps%s_stepint%s_cutoff%s-score.dict' % (seglen,nframes,step,nreps,stepint,cutoff) )
distsfile = os.path.join(outroot,'seg%s_nframes%s_step%s_nreps%s_stepint%s_cutoff%s-model_dists.dict' % (seglen,nframes,step,nreps,stepint,cutoff) )
return scorefile,distsfile
#get a dummy frame for pixel dimensions
stream = cv.CaptureFromFile(vid)
SHAPE = vidtools.array_from_stream(stream).shape
obj_train_arcs_d = {}
obj_arcs_d = {}
size_train_values_d = {}
size_values_d = {}
fol_train_values_d = {}
fol_values_d = {}
print >> sys.stderr, 'load second pass blob outlines; calculate object properties',
for i in range(nreps):
offset = (i * seglen * step) + start_offset
source_f = mousezopt.generate_outfile_names(vid, seglen, offset,
nframes, peak_max,
peak_min,
target_int_step)[1]
ols_d = eval(open(source_f).read())
ols = ([
v for k, v in ols_d.items()
if numpy.abs(cutoff - k) < target_int_step / 2
] + [[]])[0]
objs_train = [
o for o in vidtools.find_objs(ols, SHAPE, n_obj, n_obj)[0]
if len(o) > 0
]
obj_train_arcs_d[offset] = objs_train
size_train_values_d[offset] = [[
vidtools.size_of_polygon(ols[i][j]) for i, j in o
] for o in objs_train]
fol_train_values_d[offset] = [
n_obj = int(n_obj)
#get a dummy frame for pixel dimensions
stream = cv.CaptureFromFile(vid)
SHAPE = vidtools.array_from_stream(stream).shape
obj_train_arcs_d = {}
obj_arcs_d = {}
size_train_values_d = {}
size_values_d = {}
fol_train_values_d = {}
fol_values_d = {}
print >> sys.stderr, 'load second pass blob outlines; calculate object properties',
for i in range(nreps):
offset = (i*seglen*step)+start_offset
source_f = mousezopt.generate_outfile_names(vid,seglen,offset,nframes,peak_max,peak_min,target_int_step)[1]
ols_d = eval(open(source_f).read())
ols = ([v for k,v in ols_d.items() if numpy.abs(cutoff-k) < target_int_step/2]+[[]])[0]
objs_train = [o for o in vidtools.find_objs(ols,SHAPE,n_obj,n_obj)[0] if len(o) > 0]
obj_train_arcs_d[offset] = objs_train
size_train_values_d[offset] = [[vidtools.size_of_polygon(ols[i][j]) for i,j in o] for o in objs_train]
fol_train_values_d[offset] = [vidtools.fol_from_obj_arc(o,ols,SHAPE) for o in objs_train]
objs = [o for o in vidtools.find_objs(ols,SHAPE)[0] if len(o) > 0]
obj_arcs_d[offset] = objs
size_values_d[offset] = [[vidtools.size_of_polygon(ols[i][j]) for i,j in o] for o in objs]
fol_values_d[offset] = [vidtools.fol_from_obj_arc(o,ols,SHAPE) for o in objs]
print >> sys.stderr, '.',
print >> sys.stderr, 'done'
if len(reduce(lambda x,y:x+y,size_train_values_d.values())) > 0 and len(reduce(lambda x,y:x+y,fol_train_values_d.values())) > 0:
print >> sys.stderr, 'calculate scoring distributions ...' ,
