def mat_polys2cv(m,polys,dpi=80,scale=1):
pylab.close(1)
fig = pylab.figure(1,figsize=tuple(reversed([(float(i)/dpi)*scale for i in m.shape])),dpi=dpi)
ax = pylab.matshow(m,fignum=1)
for c,p in zip(iplot.subspectrum(len(polys)),polys):
ax.add_patch(matplotlib.patches.Polygon(p,fc='none',ec=c))
pylab.plot()
buf = io.BytesIO()
fig.savefig(buf,format='png')
buf.seek(0)
pi = Image.open(buf).convert('RGB')
cv_im = cv.CreateImageHeader(pi.size, cv.IPL_DEPTH_8U, 3)
cv.SetData(cv_im, pi.tostring(),pi.size[0]*3)
buf.close()
return cv_im
def draw_analysis_activity_summary(adir,fig=1,outformat=None,graph_height=0.2,show_mice=False,dawn_min=None,hill_bounds=None,skip_first=10):
'''given an analysis directory, generates a figure showing activity and grounds
if format is not None, also outputs summary figure to adir
if graph_height is not None, plots a color-coded activity graph on bottom <graph_height> portion of fig
'''
print >> sys.stderr, 'load:'
shape = eval(open(adir+'/shape.tuple').read())
if hill_bounds is None:
hill_left = 0
hill_right = shape[1]
else:
hill_left,hill_right = hill_bounds
print >> sys.stderr, '\tframes'
frames = [numpy.fromfile(f).reshape(shape) for f in sorted(glob(adir+'/*.frame'))]
actoutfs = sorted(glob(adir+'/*.newactout'))
acttermfs = sorted(glob(adir+'/*.newactterm'))
groundfs = sorted(glob(adir+'/*.ground'))
print >> sys.stderr, '\tactouts'
actouts = [eval(open(f).read()) for f in actoutfs]
print >> sys.stderr, '\tactterms'
actterms = [eval(open(f).read()) for f in acttermfs]
print >> sys.stderr, '\tgrounds'
grounds = [Util.smooth(numpy.fromfile(f,sep='\n'),10) for f in groundfs[:-1]]
hill_area = numpy.array([sum(shape[0]-g[hill_left:hill_right]) for g in grounds])
grounds = grounds[1:] #drop first ground after calculating differences
spec = iplot.subspectrum(len(actouts))
figobj = pylab.figure(fig,figsize=(9,6))
figobj.clf()
print >> sys.stderr, 'render:'
pylab.gray()
if not show_mice:
bkgd_mat = frames[-1]-(frames[0]-frames[-1])
if graph_height is not None:
h = frames[0].shape[0] * graph_height
bkgd_mat = numpy.array(list(bkgd_mat) + list(numpy.zeros((h,bkgd_mat.shape[1]))))
pylab.matshow(bkgd_mat,fignum=fig)
else:
print >> sys.stderr, 'show mice invoked; loading mice...',
mice = [eval(open(f).read()) for f in sorted(glob(adir+'/*.mice'))]
xco,yco = Util.dezip(reduce(lambda x,y: x+y,[[v for v in m.values() if v] for m in mice]))
print >> sys.stderr, 'done'
bkgd_mat = numpy.zeros(shape)
#raise NotImplementedError
if graph_height is not None:
h = frames[0].shape[0] * graph_height
bkgd_mat = numpy.array(list(bkgd_mat) + list(numpy.zeros((h,bkgd_mat.shape[1]))))
pylab.matshow(bkgd_mat,fignum=fig)
m,x,y = numpy.histogram2d(xco,yco,bins=numpy.array(shape[::-1])/5)
#the following is nasty: re-center to -0.5 bin width, log transform counts
pylab.contourf(x[1:]-(x[1]-x[0])/2,y[1:]-(y[1]-y[0])/2,numpy.log(m.transpose()+1),100)
ax = figobj.axes[0]
ax.set_xticks([])
ax.set_yticks([])
for i,poly in enumerate(actouts):
for p in poly:
if p: ax.add_patch(pylab.matplotlib.patches.Polygon(p,fc='none',ec=spec[i]))
for i, terms in enumerate(actterms):
for t in terms:
x,y = Util.dezip(t)
pylab.plot(x,y,c=spec[i])
vinf,ainf = adir.split('/analysis/')
ax.text(10,20,vinf,fontsize=8,color='w')
ax.text(10,35,ainf,fontsize=8,color='w')
if graph_height is not None:
areas = [sum([len(filter(None,vidtools.mask_from_outline(p,frames[0].shape).flat)) for p in polys]) for polys in actouts]
progs = [True and sum([vidtools.hypotenuse(*t) for t in terms]) or 0 for terms in actterms]
hill_increase = []
for v in hill_area[1:] - hill_area[:-1]:
if 0 < v:
hill_increase.append(v)
else:
hill_increase.append(0)
hill_increase = numpy.array(hill_increase)
z3 = (3*hill_increase.std()) + hill_increase.mean()
#willing to accept that a mouse kicks out at no more than 2x max observed burrow-build
max_increase = max(2*max(areas),z3)
print >> sys.stderr, 'for max ground change, z3 = %s; 2*max_areas = %s. Choose %s' % (z3,2*max(areas),max_increase)
for i,v in enumerate(hill_increase):
if v > max_increase:
hill_increase[i] = 0
h = frames[0].shape[0] * graph_height
base = frames[0].shape[0] + h/2 - 10
scale = frames[0].shape[1]/float(len(areas))
x = numpy.arange(0,frames[0].shape[1],scale)[:len(areas)]
if dawn_min:
pylab.bar(x[dawn_min],1*h/2,color='w',edgecolor='w',bottom=base)
pylab.bar(x[dawn_min],-1*h/2,color='w',edgecolor='w',bottom=base)
hours = []
for m in range(len(spec)):
if m%60==0:
hours.append(1)
else:
hours.append(0)
print len(hours)
pylab.bar(x,numpy.array(hours)*h/2,color='w',edgecolor='w',alpha=0.4,linewidth=0,bottom=base)
pylab.bar(x,-1*numpy.array(hours)*h/2,color='w',edgecolor='w',alpha=0.4,linewidth=0,bottom=base)
pylab.bar(x[skip_first:],Util.normalize(numpy.array(hill_increase[skip_first:]))*h/2,color=spec,edgecolor=spec,bottom=base)
pylab.plot(x[skip_first:],base+Util.normalize(numpy.cumsum(numpy.array(hill_increase[skip_first:])))*h/2,'--w')
for i,v in zip(x[skip_first:],Util.normalize(numpy.cumsum(numpy.array(hill_increase[skip_first:])))):
if v > 0.5:
break
pylab.plot((i,i),(base,base+(h/2)),'--w',lw=2)
pylab.bar(x[skip_first:],-1*Util.normalize(numpy.array(areas[skip_first:]))*h/2,color=spec,edgecolor=spec,bottom=base)
pylab.plot(x[skip_first:],base-Util.normalize(numpy.cumsum(numpy.array(areas[skip_first:])))*h/2,'--w')
for i,v in zip(x[skip_first:],Util.normalize(numpy.cumsum(numpy.array(areas[skip_first:])))):
if v > 0.5:
break
pylab.plot((i,i),(base,base-(h/2)),'--w',lw=2)
pylab.text(5,base+h/2,'%0.1f' % max(hill_increase[skip_first:]),color='w')
pylab.text(5,base-h/2,'%0.1f' % max(areas[skip_first:]),color='w')
#'''
spec.reverse() #flip spectrum to draw grounds back-to-front
for i, g in enumerate(grounds[::-1]):
pylab.plot(numpy.arange(hill_left,hill_right),g[hill_left:hill_right],c=spec[i])
spec.reverse() #flip back
#'''
pylab.plot()
pylab.ylim(1.2*shape[0],0)
pylab.xlim(0,shape[1])
if outformat is not None:
print >> sys.stderr, 'save as '+adir+'/activity_summary.'+outformat
figobj.savefig(adir+'/activity_summary.'+outformat)
def draw_rainbowtron_opencv(analysis_dir,fig=1,outformat='pdf',vid_fps=30,graph_height=0.2, \
bkgd_data='fullmice',spectrum_len=None,dawn_min=None,activity_percentile_plot=0.5, \
hill_bounds=None,skip_first=0,hsl_per_min=2,set_hill_max=None,set_dig_max=None, \
post_summarize_analysis=True,outfile_prefix=None, \
ground_step_coeff=1,max_pix_mult=None,max_obj=None,max_dist_mult=None,hill_win_coeff=None,write_yval=False, \
hill_max_cumul_digscale=None,dig_max_cumul_digscale=None):
'''given a summarize_segment_opencv.py analysis directory (i.e. tarballs of analysis results by half-segment-length)
generates a figure showing burrowing activity and sand-hill accumulation
if outformat is not None, save figure to analysis_dir
if vidformat is not None, record rainbowtron over time to analysis_dir
bkgd_data: data to display as background for image. one of ["fullmice","flatmice","hillchange"]
if post_summarize_analysis is True, look for ground and activity from vidtools.find_ground_over_frames_matrix; vidtools.find_burrow_over_frames_matrix
HACK: if bkgd_data is NOT "fullmice", the mouse size vector will not be loaded, and a mouse_radius of 10 will be used
'''
if outfile_prefix is None:
outfile_prefix = analysis_dir.rstrip('/').split('/')[-4].split('-')[0] + '-' + ('_'.join(analysis_dir.rstrip('/').split('/')[-3].split('_')[-2:]))
outfile_prefix = '%s-hillmax%s-digmax%s-gsc%s-mpm%s-mo%s-mdm%s-hwc%s-hmcd%s-dmcd%s-bkgd_%s' % \
(outfile_prefix, set_hill_max,set_dig_max, \
ground_step_coeff,max_pix_mult,max_obj,max_dist_mult,hill_win_coeff,hill_max_cumul_digscale,dig_max_cumul_digscale, \
bkgd_data)
#will get overwritten later if possible
mouse_radius = 10
print >> sys.stderr, 'load:'
SHAPE = eval(open(analysis_dir+'/SHAPE').read())
config = eval(open(os.path.split(os.path.split(analysis_dir.rstrip('/'))[0])[0]+'-config.dict').read())
if hill_bounds is None:
hill_left = 0
hill_right = SHAPE[1]
elif hill_bounds == 'config':
hill_left,hill_right = config['hill_bounds']
else:
hill_left,hill_right = hill_bounds
print >> sys.stderr, '\ttar handles'
tars_d = dict([(os.path.basename(f),tarfile.open(f)) for f in glob(os.path.join(analysis_dir,'*.tar'))])
print >> sys.stderr, '\tframes'
frames = Util.tar2ar_all(tars_d['segavgs.tar'],SHAPE)
if bkgd_data == 'hillchange':
bkgd_mat = frames[-1]-(frames[0]-frames[-1])
else:
if bkgd_data == 'flatmice':
print >> sys.stderr, '\tmouse masks'
mouse_hists = Util.tar2ar_all(tars_d['mousemasks.tar'],SHAPE,bool)
elif bkgd_data == 'fullmice':
print >> sys.stderr, '\tmouse locations'
hist2d_tarf = os.path.join(analysis_dir,'mousehists-mpm%s-mo%s-mdm%s.tar' % (max_pix_mult,max_obj,max_dist_mult) )
outlines_tarf = os.path.join(analysis_dir,'miceols.tar')
# this will save tons of time...eventually...
mouse_hists,svec = load_cached_mouse_hist2d_segments(hist2d_tarf,outlines_tarf,SHAPE,check_lengths=True,return_sizes=True, \
max_pix_mult=max_pix_mult,max_obj=max_obj,max_dist_mult=max_dist_mult)
#mouse_hists = load_mouse_hist2d_segments(tars_d['miceols.tar'],SHAPE)
mouse_radius = math.sqrt(numpy.mean(svec)/math.pi)
else:
errstr = 'bkgd_data must be one of ["hillchange","flatmice","fullmice"]; got %s' % bkgd_data
raise ValueError, errstr
bkgd_mat = numpy.log1p(reduce(lambda x,y:x+numpy.array(y,dtype=int), mouse_hists))
print >> sys.stderr, '\tgrounds'
if post_summarize_analysis:
grounds_postsummary_tarbase = 'grounds_postsummary-gsc%s.tar' % (ground_step_coeff)
if grounds_postsummary_tarbase in tars_d:
grounds = Util.tar2obj_all(tars_d[grounds_postsummary_tarbase],lambda t:numpy.array(eval(t),dtype=float))
print >> sys.stderr, 'loaded %s grounds' % len(grounds)
else:
grounds_postsummary_tarf = os.path.join(analysis_dir,grounds_postsummary_tarbase)
grounds = vidtools.find_ground_over_frames_matrix(frames,config['ground_anchors'],config['end_ground_anchors'],max_step_limit=mouse_radius*ground_step_coeff)
#write grounds so future runs can load cached
tarh = tarfile.open(grounds_postsummary_tarf,'w')
for g,n in zip(grounds,sorted(tars_d['grounds.tar'].getnames())):
Util.obj2tar(g,n,tarh)
tarh.close()
grounds = map(numpy.array,grounds)
else:
grounds = [Util.smooth(g,10) for g in Util.tar2obj_all(tars_d['grounds.tar'],lambda t:numpy.array(eval(t),dtype=float))]
print >> sys.stderr, '\tactouts'
if post_summarize_analysis:
newact_ols_postsummary_tarbase = 'newact_ols_postsummary-gsc%s-mpm%s-mo%s-mdm%s.tar' % \
(ground_step_coeff,max_pix_mult,max_obj,max_dist_mult)
pd_ols_postsummary_tarbase = 'pd_ols_postsummary-gsc%s-mpm%s-mo%s-mdm%s.tar' % \
(ground_step_coeff,max_pix_mult,max_obj,max_dist_mult)
newact_ols_postsummary_tarf = os.path.join(analysis_dir,newact_ols_postsummary_tarbase)
pd_ols_postsummary_tarf = os.path.join(analysis_dir,pd_ols_postsummary_tarbase)
if newact_ols_postsummary_tarbase in tars_d and pd_ols_postsummary_tarbase in tars_d:
actouts = Util.tar2obj_all(tars_d[newact_ols_postsummary_tarbase])
pdols = Util.tar2obj_all(tars_d[pd_ols_postsummary_tarbase])
print >> sys.stderr, 'loaded %s activity outlines' % len(actouts)
print >> sys.stderr, 'loaded %s predug outlines' % len(pdols)
else:
if bkgd_data == 'hillchange':
mousemasks = Util.tar2ar_all(tars_d['mousemasks.tar'],SHAPE,bool)
else:
mousemasks = [numpy.array(m,dtype=bool) for m in mouse_hists]
pdmask,pdols,prevactols,actouts = vidtools.find_burrow_over_frames_matrix(frames,mousemasks,grounds,mouse_radius,predug_poly=config.get('predug',None))
na_tarh = tarfile.open(newact_ols_postsummary_tarf,'w')
pd_tarh = tarfile.open(pd_ols_postsummary_tarf,'w')
for na,pd,n in zip(actouts,pdols,sorted(tars_d['newact_ols.tar'].getnames())):
Util.obj2tar(na,n,na_tarh)
Util.obj2tar(pd,n,pd_tarh)
na_tarh.close()
pd_tarh.close()
else:
actouts = Util.tar2obj_all(tars_d['newact_ols.tar'])
if hill_win_coeff is None:
hill_area = numpy.array([sum(SHAPE[0]-g[hill_left:hill_right]) for g in grounds])
hill_increase = []
for v in hill_area[1:] - hill_area[:-1]:
if 0 < v:
hill_increase.append(v)
else:
hill_increase.append(0)
else:
#ADD HILL AREA WINDOW CALC
win = mouse_radius * hill_win_coeff
hill_increase = []
for i,g in enumerate(grounds[1:]):
diff_vec = grounds[i] - g
hi,win_start = max([(sum(diff_vec[x:x+win]),x) for x in range(len(diff_vec))])
hill_increase.append(hi)
hill_increase = numpy.array(hill_increase)
grounds = grounds[1:] #drop first ground after calculating differences
# takes second (1-th) actout since first is always blank
areas = [sum([len(filter(None,vidtools.mask_from_outline(p,frames[0].shape).flat)) for p in polys]) for polys in actouts[1:]]
#progs = [True and sum([vidtools.hypotenuse(*t) for t in terms]) or 0 for terms in actterms]
#write hill increase and dig areas
areas_out = os.path.join(analysis_dir,outfile_prefix+'-dig_areas.list')
hills_out = os.path.join(analysis_dir,outfile_prefix+'-hill_increases.list')
open(areas_out,'w').write(list(areas).__repr__())
open(hills_out,'w').write(list(hill_increase).__repr__())
#PLOT
if hill_max_cumul_digscale:
hill_scale_factor = sum(hill_increase)/float(hill_max_cumul_digscale)
else:
hill_scale_factor = 1
if dig_max_cumul_digscale:
dig_scale_factor = sum(areas)/float(dig_max_cumul_digscale)
else:
dig_scale_factor = 1
if spectrum_len is None:
spectrum_len = len(actouts)
spec = iplot.subspectrum(spectrum_len,reversed=True,noUV=True)
figobj = pylab.figure(fig,figsize=(9,6))
figobj.clf()
print >> sys.stderr, 'render:'
pylab.gray()
if graph_height is not None:
h = frames[0].shape[0] * graph_height
bkgd_mat = numpy.array(list(bkgd_mat) + list(numpy.zeros((h,bkgd_mat.shape[1]))))
pylab.matshow(bkgd_mat,fignum=fig)
ax = figobj.axes[0]
ax.set_xticks([])
ax.set_yticks([])
for i,poly in enumerate(actouts):
for p in poly:
if p: ax.add_patch(pylab.matplotlib.patches.Polygon(p,fc='none',ec=spec[i]))
if post_summarize_analysis and len(config.get('predug',[])) > 0:
for i,poly in enumerate(pdols):
for p in poly:
if p: ax.add_patch(matplotlib.patches.Polygon(p,ec='gray',fc='none',hatch='/'))
vinf,ainf = analysis_dir.split('/analysis/')
ax.text(10,20,vinf,fontsize=8,color='w')
ax.text(10,35,ainf,fontsize=8,color='w')
if graph_height is not None:
z3 = (3*hill_increase.std()) + hill_increase.mean()
#willing to accept that a mouse kicks out at no more than 2x max observed burrow-build?
max_increase = max(2*max(areas),z3)
print >> sys.stderr, 'for max ground change, z3 = %s; 2*max_areas = %s. Choose %s' % (z3,2*max(areas),max_increase)
for i,v in enumerate(hill_increase):
if v > max_increase:
hill_increase[i] = 0
h = frames[0].shape[0] * graph_height
base = frames[0].shape[0] + h/2 - 10
scale = frames[0].shape[1]/float(len(areas))
x = numpy.arange(0,frames[0].shape[1],scale)[:len(areas)]
print len(x),len(hill_increase)
if dawn_min:
pylab.bar(x[dawn_min],1*h/2,color='w',edgecolor='w',bottom=base)
pylab.bar(x[dawn_min],-1*h/2,color='w',edgecolor='w',bottom=base)
hours = []
for m in range(min(len(spec),len(areas))):
if m%(60*hsl_per_min)==0:
hours.append(1)
else:
hours.append(0)
print len(hours)
pylab.bar(x,numpy.array(hours)*h/2,color='w',edgecolor='w',alpha=0.4,linewidth=0,bottom=base)
pylab.bar(x,-1*numpy.array(hours)*h/2,color='w',edgecolor='w',alpha=0.4,linewidth=0,bottom=base)
pylab.bar(x[skip_first:],-1*Util.normalize(numpy.array(hill_increase[skip_first:]),set_max=set_hill_max)*h/2,color=spec,edgecolor=spec,bottom=base)
pylab.plot(x[skip_first:],base-Util.normalize(numpy.cumsum(numpy.array(hill_increase[skip_first:])))*h/2*hill_scale_factor,'--w')
for i,v in zip(x[skip_first:],Util.normalize(numpy.cumsum(numpy.array(hill_increase[skip_first:])))):
if activity_percentile_plot and v > activity_percentile_plot:
break
pylab.plot((i,i),(base,base-(h/2)),'--w',lw=2)
pylab.bar(x[skip_first:],Util.normalize(numpy.array(areas[skip_first:]),set_max=set_dig_max)*h/2,color=spec,edgecolor=spec,bottom=base)
pylab.plot(x[skip_first:],base+Util.normalize(numpy.cumsum(numpy.array(areas[skip_first:])))*h/2*dig_scale_factor,'--w')
for i,v in zip(x[skip_first:],Util.normalize(numpy.cumsum(numpy.array(areas[skip_first:])))):
if activity_percentile_plot and v > activity_percentile_plot:
break
pylab.plot((i,i),(base,base+(h/2)),'--w',lw=2)
if write_yval:
pylab.text(5,base-h/2,'hill %0.1f' % max(hill_increase[skip_first:]),color='w')
pylab.text(5,base+h/2,'dig %0.1f' % max(areas[skip_first:]),color='w')
#'''
for i, g in enumerate(grounds):
pylab.plot(numpy.arange(hill_left,hill_right),g[hill_left:hill_right],c=spec[i])
#'''
pylab.plot()
pylab.ylim((1+graph_height)*SHAPE[0],0)
pylab.xlim(0,SHAPE[1])
if outformat is not None:
out_img = os.path.join(analysis_dir,outfile_prefix+'-activity_summary.'+outformat)
print >> sys.stderr, 'save as %s' % out_img
figobj.savefig(out_img)
#VIDEO!
spec = iplot.subspectrum(spectrum_len,noUV=True)
if vid_fps:
import cv
vidout = os.path.join(analysis_dir,outfile_prefix+'-activity_summary.avi')
if os.path.exists(vidout): os.unlink(vidout)
pixdim = tuple(reversed(SHAPE))
pixdim = tuple(numpy.array(pixdim)*2)
vidwriter = cv.CreateVideoWriter(vidout , cv.FOURCC('x','v','i','d'), vid_fps, pixdim,1)
cols_polys = []
cols_hatch_polys = []
cols_lines = []
hourbar_kwargs = {'alpha':0.4,'linewidth':0}
bars = []
bars.append(('w',(x,numpy.array(hours)*h/2,base,hourbar_kwargs)))
bars.append(('w',(x,-1*numpy.array(hours)*h/2,base,hourbar_kwargs)))
for vfi,m in enumerate(mouse_hists[1:]):
if graph_height is not None:
h = frames[0].shape[0] * graph_height
m = numpy.array(list(m) + list(numpy.zeros((h,m.shape[1]))))
if len(actouts[vfi]):
cols_polys.extend(zip([spec[vfi]]*len(actouts[vfi]),actouts[vfi]))
if post_summarize_analysis and len(config.get('predug',[])) > 0 and len(pdols[vfi]):
cols_hatch_polys.extend(zip(['gray']*len(pdols[vfi]),['/']*len(pdols[vfi]),pdols[vfi]))
cols_lines.append((spec[vfi],(numpy.arange(hill_left,hill_right),grounds[vfi][hill_left:hill_right])))
this_bars = []
#pylab.bar(x[skip_first:],-1*Util.normalize(numpy.array(hill_increase[skip_first:]))*h/2,color=spec,edgecolor=spec,bottom=base)
this_bars.append((spec[:vfi],(x[:vfi],(-1*Util.normalize(numpy.array(hill_increase))*h/2)[:vfi],base,{})))
#pylab.plot(x[skip_first:],base-Util.normalize(numpy.cumsum(numpy.array(hill_increase[skip_first:])))*h/2,'--w')
hi_line = [('w',(x[:vfi],(base-Util.normalize(numpy.cumsum(numpy.array(hill_increase)))*h/2)[:vfi]))]
#pylab.bar(x[skip_first:],Util.normalize(numpy.array(areas[skip_first:]))*h/2,color=spec,edgecolor=spec,bottom=base)
this_bars.append((spec[:vfi],(x[:vfi],(Util.normalize(numpy.array(areas))*h/2)[:vfi],base,{})))
#pylab.plot(x[skip_first:],base+Util.normalize(numpy.cumsum(numpy.array(areas[skip_first:])))*h/2,'--w')
ba_line = [('w',(x[:vfi],(base+Util.normalize(numpy.cumsum(numpy.array(areas)))*h/2)[:vfi]))]
cv_im = vidtools.mat_polys2cv(m,cols_polys,hatchpolys=cols_hatch_polys,lines=cols_lines+hi_line+ba_line,bars=bars+this_bars,scale=2)
nll = cv.WriteFrame(vidwriter,cv_im)
print >> sys.stderr, '\r\twrite video frame %s' % vfi,
def draw_reanalysis_activity_summary(re_adir,source_adir,fig=1,outformat=None):
'''given a reanalysis directory and a source directory, generates a figure showing previous and concurrent activity from source
and activity found in reanalysis
if format is not None, also outputs to re_adir
'''
print >> sys.stderr, 'load:'
shape = eval(open(re_adir+'shape.tuple').read())
frames = [numpy.fromfile(f).reshape(shape) for f in sorted(glob(re_adir+'/*.frame'))]
s,e,fps = re.search('/(\d+)-(\d+)/(\d+)fps/',re_adir).groups()
source_analysis_window,source_fps = re.search('/([\d\w-]+)/(\d+)fps/',source_adir).groups()
offset_fr = 0
if '-' in source_analysis_window:
start = source_analysis_window.split('-')[0]
if start != 'start':
offset_fr = int(start)*int(source_fps)
win = [int(fps)*int(n) for n in s,e]
preactoutfs = sorted(glob(source_adir+'/*.preactout'))
minactoutfs = sorted(glob(source_adir+'/*.newactout'))
newacttermfs = sorted(glob(re_adir+'/*.newactterm'))
preactouts = eval(open(vidtools.get_files_by_win(preactoutfs,win,offset_fr)[0]).read())
minactouts,term,mask = vidtools.merge_polygons(reduce(lambda x,y:x+y,filter(None,[eval(open(f).read()) for f in vidtools.get_files_by_win(minactoutfs,win,offset_fr)])),shape,(0,0))
secactouts = [eval(open(f).read()) for f in sorted(glob(re_adir+'*.newactout'))]
newactterms = [eval(open(f).read()) for f in newacttermfs]
spec = iplot.subspectrum(len(secactouts))
figobj = pylab.figure(fig,figsize=(9,6))
figobj.clf()
print >> sys.stderr, 'render:'
pylab.matshow(frames[-1]-(frames[0]-frames[-1]),fignum=fig)
ax = figobj.axes[0]
ax.set_xticks([])
ax.set_yticks([])
for p in preactouts:
if p: ax.add_patch(pylab.matplotlib.patches.Polygon(p,fc='none',ec='k'))
for p in minactouts:
if p: ax.add_patch(pylab.matplotlib.patches.Polygon(p,fc='none',ec='w',lw=2))
print >>sys.stderr, 'len spec %s len secactouts %s' % (len(spec), len(secactouts))
for i,poly in enumerate(secactouts):
for p in poly:
if p: ax.add_patch(pylab.matplotlib.patches.Polygon(p,fc='none',ec=spec[i]))
for i, terms in enumerate(newactterms):
for t in terms:
x,y = Util.dezip(t)
pylab.plot(x,y,c=spec[i])
pylab.gray()
ax.text(10,20,'%s-%s %s-%s' % (win[0],win[1],Util.hms_from_sec(win[0]/int(fps)),Util.hms_from_sec(win[1]/int(fps))),fontsize=16,color='w')
pylab.plot()
if outformat is not None:
print >> sys.stderr, 'save as '+re_adir+'/summary.'+outformat
figobj.savefig(re_adir+'/summary.'+outformat)
nll = mousemasks.pop(0)
segavgs.append(last_avg)
if len(segavgs) > 3:
nll = segavgs.pop(0)
segmasked.append(last_masked)
if len(segmasked) > 3:
nll = segmasked.pop(0)
prelast_avg = last_avg
prelast_mm = prelast_mm + last_mm
if opts.antfarm_config:
grounds.append(last_ground)
#done handling data
#DRAW FRAME RESULTS
if opts.video_suffix:
if opts.antfarm_config:
this_color = iplot.subspectrum(analyze_hsls+1)[i]
if i > 0:
g_col_li = [('k',zip(*list(enumerate(Util.smooth(grounds[0],10)))))]
else:
g_col_li = []
g_col_li.append((this_color,zip(*list(enumerate(Util.smooth(grounds[-1],10))))))
last_frames_objs_ols = vidtools.ols_in_interval(frames_offset-hsl-1,frames_offset,min_arc_score,ols,ols_offset, \
Util.merge_dictlist([objs,to_retire_objs]), \
Util.merge_dictlist([objs_sizes,to_retire_objs_sizes]), \
Util.merge_dictlist([objs_fols,to_retire_objs_fols]), \
size_h, size_bins, fol_h, fol_bins,True)
for vfi,m in enumerate(last_frames):
if opts.antfarm_config:
cv_im = vidtools.mat_polys2cv(m,zip(iplot.subspectrum(6)[1:],list(reversed(last_frames_objs_ols[vfi][:5])))+zip(['k']*len(prevol),prevol),zip([this_color]*len(digol),['\\']*len(digol),digol),g_col_li,80,2)
nll = mousemasks.pop(0)
segavgs.append(last_avg)
if len(segavgs) > 3:
nll = segavgs.pop(0)
segmasked.append(last_masked)
if len(segmasked) > 3:
nll = segmasked.pop(0)
prelast_avg = last_avg
prelast_mm = prelast_mm + last_mm
if opts.antfarm_config:
grounds.append(last_ground)
#done handling data
#DRAW FRAME RESULTS
if opts.video_suffix:
if opts.antfarm_config:
this_color = iplot.subspectrum(analyze_hsls + 1)[i]
if i > 0:
g_col_li = [
('k',
zip(*list(enumerate(Util.smooth(grounds[0], 10)))))
]
else:
g_col_li = []
g_col_li.append(
(this_color,
zip(*list(enumerate(Util.smooth(grounds[-1], 10))))))
last_frames_objs_ols = vidtools.ols_in_interval(frames_offset-hsl-1,frames_offset,min_arc_score,ols,ols_offset, \
Util.merge_dictlist([objs,to_retire_objs]), \
Util.merge_dictlist([objs_sizes,to_retire_objs_sizes]), \
Util.merge_dictlist([objs_fols,to_retire_objs_fols]), \
tunnels['MR1.1'] = ginput(2,0) slope(*tunnels['MR1.1']) degrees(arctan(slope(*tunnels['MR1.1']))) degrees(arctan(slope(*tunnels['MR1.1']))) - degrees(arctan(slope(*tunnels['MLhill']))) degrees(arctan(slope(*tunnels['MR1.1']))) - degrees(arctan(slope(*horiz))) #?raw_input gray() import Util x,y = Util.dezip(tunnels['MLhill']) x y plot(x,y,'w--',lw=2) x,y = Util.dezip(tunnels['MR1.1']) import iplot spec = iplot(10) spec = iplot.subspectrum(10) plot(x,y,c=spec[0],lw=2) tunnels['MR1.2'] = ginput(5,0) x,y = Util.dezip(tunnels['MR1.2']) plot(x,y,c=spec[1],lw=2) degrees(arctan(slope(*tunnels['MR1.2'][:2]))) - degrees(arctan(slope(*tunnels['MR1.1']))) degrees(arctan(slope(*tunnels['MR1.2'][:2]))) - degrees(arctan(slope(*tunnels['MR1.1']))) degrees(arctan(slope(*tunnels['MR1.1']))) - degrees(arctan(slope(*horiz))) degrees(arctan(slope(*tunnels['MR1.2'][:2]))) - degrees(arctan(slope(*horiz))) tunnels['MR1.3'] = ginput(2,0) tunnels['MR1.4'] = ginput(4,0) tunnels['MR1.5'] = ginput(2,0) x,y = Util.dezip(tunnels['MR1.3']) plot(x,y,c=spec[2],lw=2) x,y = Util.dezip(tunnels['MR1.4']) plot(x,y,c=spec[3],lw=2)
def run_full_plot(frags,clust_members,indivs_by_pool,pool,samp_sd,sf1=0.95,sf2=0.9):
from pylab import plot, scatter, subplot, savefig, clf, title
import iplot, numpy
matched_reads = []
losd_reads = []
hisd_reads = []
for ind in indivs_by_pool[pool]:
numreads = sum(clust_members[ind].values())
print >> sys.stderr, ind,numreads
mr = simulate_coverage(frags,300, samp_sd,numreads,0,1000)
matched_reads.append(mr)
mr = simulate_coverage(frags,300, samp_sd-4,numreads,0,1000)
losd_reads.append(mr)
mr = simulate_coverage(frags,300, samp_sd+4,numreads,0,1000)
hisd_reads.append(mr)
matched_reads_d = dict([((i,sum(mr.values())),mr) for i,mr in enumerate(matched_reads)])
losd_reads_d = dict([((i,sum(mr.values())),mr) for i,mr in enumerate(losd_reads)])
hisd_reads_d = dict([((i,sum(mr.values())),mr) for i,mr in enumerate(hisd_reads)])
obs_reads = [clust_members[ind] for ind in indivs_by_pool[pool]]
obs_reads_d = dict([((i,sum(mr.values())),mr) for i,mr in enumerate(obs_reads)])
clf()
subplot(2,2,1)
plotcols = iplot.subspectrum(len(indivs_by_pool[pool]))
for col,ind,mr in zip(plotcols, indivs_by_pool[pool], matched_reads):
plot(numpy.log1p(sorted(clust_members[ind].values(),reverse=True))[10:100000],color=col)
plot(numpy.log1p(sorted(mr.values(),reverse=True))[10:100000],color=col,ls='--',lw=1)
title(pool)
subplot(2,2,2)
sf = 1
x,y = Util.dezip(sorted([(sum(v.values()),numpy.mean([i for i in v.values() if i>1])) for v in matched_reads]))
plot(x,sf*numpy.array(y),'k')
x,y = Util.dezip(sorted([(sum(v.values()),numpy.mean([i for i in v.values() if i>1])) for v in losd_reads]))
plot(x,sf*numpy.array(y),'k:')
x,y = Util.dezip(sorted([(sum(v.values()),numpy.mean([i for i in v.values() if i>1])) for v in hisd_reads]))
plot(x,sf*numpy.array(y),'k--')
x,y = Util.dezip([(sum(clust_members[ind].values()),numpy.mean([i for i in clust_members[ind].values() if i>1])) for ind in indivs_by_pool[pool]])
scatter(x,y,c='none',lw=2)
subplot(2,2,3)
sf = sf1
x,y = Util.dezip([(sum(clust_members[ind].values()),len([i for i in clust_members[ind].values() if i >= 4])) for ind in indivs_by_pool[pool]])
scatter(x,y,c='none',edgecolor='r',lw=2)
x,y = Util.dezip(sorted([(sum(v.values()),len([i for i in v.values() if i >= 4])) for v in matched_reads]))
plot(x,sf*numpy.array(y),'r')
x,y = Util.dezip(sorted([(sum(v.values()),len([i for i in v.values() if i >= 4])) for v in losd_reads]))
plot(x,sf*numpy.array(y),'r:')
x,y = Util.dezip(sorted([(sum(v.values()),len([i for i in v.values() if i >= 4])) for v in hisd_reads]))
plot(x,sf*numpy.array(y),'r--')
x,y = Util.dezip([(sum(clust_members[ind].values()),len([i for i in clust_members[ind].values() if i >= 7])) for ind in indivs_by_pool[pool]])
scatter(x,y,c='none',edgecolor='g',lw=2)
x,y = Util.dezip(sorted([(sum(v.values()),len([i for i in v.values() if i >= 7])) for v in matched_reads]))
plot(x,sf*numpy.array(y),'g')
x,y = Util.dezip(sorted([(sum(v.values()),len([i for i in v.values() if i >= 7])) for v in losd_reads]))
plot(x,sf*numpy.array(y),'g:')
x,y = Util.dezip(sorted([(sum(v.values()),len([i for i in v.values() if i >= 7])) for v in hisd_reads]))
plot(x,sf*numpy.array(y),'g--')
x,y = Util.dezip([(sum(clust_members[ind].values()),len([i for i in clust_members[ind].values() if i >= 10])) for ind in indivs_by_pool[pool]])
scatter(x,y,c='none',edgecolor='b',lw=2)
x,y = Util.dezip(sorted([(sum(v.values()),len([i for i in v.values() if i >= 10])) for v in matched_reads]))
plot(x,sf*numpy.array(y),'b')
x,y = Util.dezip(sorted([(sum(v.values()),len([i for i in v.values() if i >= 10])) for v in losd_reads]))
plot(x,sf*numpy.array(y),'b:')
x,y = Util.dezip(sorted([(sum(v.values()),len([i for i in v.values() if i >= 10])) for v in hisd_reads]))
plot(x,sf*numpy.array(y),'b--')
subplot(2,2,4)
sf = sf2
shared_mat = calc_shared_matrix(matched_reads_d,4)
lo_shared_mat = calc_shared_matrix(losd_reads_d,4)
hi_shared_mat = calc_shared_matrix(hisd_reads_d,4)
obs_shared_mat = calc_shared_matrix(obs_reads_d,4)
x,y = Util.dezip(sorted([(k[1],numpy.mean(obs_shared_mat[k[0]])) for k in obs_reads_d.keys()]))
scatter(x,y,c='none',edgecolor='r',lw=2)
x,y = Util.dezip(sorted([(k[1],numpy.mean(shared_mat[k[0]])) for k in matched_reads_d.keys()]))
plot(x,sf*numpy.array(y),'r')
x,y = Util.dezip(sorted([(k[1],numpy.mean(lo_shared_mat[k[0]])) for k in losd_reads_d.keys()]))
plot(x,sf*numpy.array(y),'r:')
x,y = Util.dezip(sorted([(k[1],numpy.mean(hi_shared_mat[k[0]])) for k in hisd_reads_d.keys()]))
plot(x,sf*numpy.array(y),'r--')
shared_mat = calc_shared_matrix(matched_reads_d,7)
lo_shared_mat = calc_shared_matrix(losd_reads_d,7)
hi_shared_mat = calc_shared_matrix(hisd_reads_d,7)
obs_shared_mat = calc_shared_matrix(obs_reads_d,7)
x,y = Util.dezip(sorted([(k[1],numpy.mean(obs_shared_mat[k[0]])) for k in obs_reads_d.keys()]))
scatter(x,y,c='none',edgecolor='g',lw=2)
x,y = Util.dezip(sorted([(k[1],numpy.mean(shared_mat[k[0]])) for k in matched_reads_d.keys()]))
plot(x,sf*numpy.array(y),'g')
x,y = Util.dezip(sorted([(k[1],numpy.mean(lo_shared_mat[k[0]])) for k in losd_reads_d.keys()]))
plot(x,sf*numpy.array(y),'g:')
x,y = Util.dezip(sorted([(k[1],numpy.mean(hi_shared_mat[k[0]])) for k in hisd_reads_d.keys()]))
plot(x,sf*numpy.array(y),'g--')
shared_mat = calc_shared_matrix(matched_reads_d,10)
lo_shared_mat = calc_shared_matrix(losd_reads_d,10)
hi_shared_mat = calc_shared_matrix(hisd_reads_d,10)
obs_shared_mat = calc_shared_matrix(obs_reads_d,10)
x,y = Util.dezip(sorted([(k[1],numpy.mean(obs_shared_mat[k[0]])) for k in obs_reads_d.keys()]))
scatter(x,y,c='none',edgecolor='b',lw=2)
x,y = Util.dezip(sorted([(k[1],numpy.mean(shared_mat[k[0]])) for k in matched_reads_d.keys()]))
plot(x,sf*numpy.array(y),'b')
x,y = Util.dezip(sorted([(k[1],numpy.mean(lo_shared_mat[k[0]])) for k in losd_reads_d.keys()]))
plot(x,sf*numpy.array(y),'b:')
x,y = Util.dezip(sorted([(k[1],numpy.mean(hi_shared_mat[k[0]])) for k in hisd_reads_d.keys()]))
plot(x,sf*numpy.array(y),'b--')
