def medianposture(steps,info,cropsize=(300,300),ax=None):
if ax is None:
fig = plt.figure()
ax = fig.gca()
xhead = activitytables.flipleftwards(steps.xhead,steps.side)
median = steps[xhead == xhead.median()].iloc[-1,:]
stepindex = 4 if median.side == 'leftwards' else 3
stepcenter = steprois_crop.center[stepindex]
info = info.ix[median.name[:2],:]
videopath = datapath.relativepath(info,'front_video.avi')
background = video.readsinglebackground(videopath,median.frame)
background = cv2.cvtColor(background,cv2.cv.CV_GRAY2BGR)
background *= 2.1
frame = video.readsingleframe(videopath,median.frame,segmented=True)
_,mask = cv2.threshold(frame,3,255,cv2.cv.CV_THRESH_BINARY)
mask = mask.astype(np.bool)
frame = cv2.cvtColor(frame,cv2.cv.CV_GRAY2BGR)
frame *= 2.1
frame[mask,...] += background[mask,...]
fliprois = RoiSet(steprois_pixels.rois,flipxy=True)
pts = np.array(fliprois.rois[stepindex])
cv2.polylines(background,[pts],True,(0,0,255),2)
background[mask,...] = frame[mask,...]
frame = background
frame = imgproc.croprect(stepcenter,cropsize,frame)
ax.imshow(frame)
x,y = _cmtopixel_(median.xhead,median.yhead)
x,y = _pixelcrop_(x,y,stepcenter,cropsize)
ax.scatter(x,y,color='r')
ax.set_axis_off()
def posturemean(steps,color='b',label=None,ax=None):
if ax is None:
fig = plt.figure()
ax = fig.gca()
xhead = activitytables.flipleftwards(steps.xhead,steps.side)
xmean = xhead.mean()
ymean = steps.yhead.mean()
xerr = xhead.sem()
yerr = steps.yhead.sem()
ax.scatter(xmean,ymean,color=color)
ax.errorbar(xmean,ymean,xerr=xerr,yerr=yerr,ecolor=color)
if label is not None:
ax.annotate(label,xy=(xmean,ymean),textcoords='offset points')
ax.set_xlabel('progression (cm)')
ax.set_ylabel('height (cm)')
from datapath import jumpers, lesionshamcache, stepfeatures_key
from infotables import names, control, lesion, smalllesion, anylesion
# Load data
stable = '(3 <= session < 5)'
unstable = '(9 <= session < 11)'
nonjumpers = str.format("subject not in {0}", jumpers)
sessions = str.format("{0} or {1}", stable, unstable)
info = pd.read_hdf(lesionshamcache, info_key)
lesions = str.format("subject in {0}", list(names(lesion(info))))
smalllesions = str.format("subject in {0}", list(names(smalllesion(info))))
alllesions = str.format("subject in {0}", list(names(anylesion(info))))
controls = str.format("subject in {0}", list(names(control(info))))
steps = pd.read_hdf(lesionshamcache, stepfeatures_key).query(nonjumpers)
steps.xhead_speed[steps.side == 'leftwards'] *= -1
steps.xhead = flipleftwards(steps.xhead, steps.side)
steps = steps.query(sessions)
# Plot functions
rangex = (steps.xhead_speed.min(), steps.xhead_speed.max())
rangey = (steps.yhead_speed.min(), steps.yhead_speed.max())
def histogramcomparison(steps, title):
axes = scatterhistaxes()
axScatter, axHistx, axHisty = axes
speedhistogram(steps.query(stable), rangex, rangey, color='b', axes=axes)
speedhistogram(steps.query(unstable), rangex, rangey, color='r', axes=axes)
axScatter.legend(['stable', 'unstable'], loc=2)
ntrials = len(steps.query(stable)) + len(steps.query(unstable))
axHistx.set_title(str.format("n = {0} trials", ntrials))
selection = str.format("subject in {0}", [name])
scr = cr.query(selection)
stablebias, unstablebias = posturebias(scr, n=bias)
stablebias = getballistictrials(stablebias)
unstablebias = getballistictrials(unstablebias)
if len(stablebias) == 0 or len(unstablebias) == 0:
continue
# Select data
stablebias = stablebias.rename(columns={'index': 'crossing'})
unstablebias = unstablebias.rename(columns={'index': 'crossing'})
stablebias.set_index('crossing', append=True, inplace=True)
unstablebias.set_index('crossing', append=True, inplace=True)
scract = cract.join(stablebias, how='inner', rsuffix='R')
ucract = cract.join(unstablebias, how='inner', rsuffix='R')
scract.xhead = flipleftwards(scract.xhead, scract.side)
ucract.xhead = flipleftwards(ucract.xhead, ucract.side)
sb_S = scract.query('stepstate3')
sb_U = scract.query('not stepstate3')
ub_S = ucract.query('stepstate3')
ub_U = ucract.query('not stepstate3')
# Plot data
name = namemap[name]
fig, (ax1, ax2) = plt.subplots(1, 2)
colors = ['b', 'cyan', 'orange', 'r']
averagetimeseries(sb_S, color=colors[0], ax=ax1, alpha=0.25)
averagetimeseries(sb_U, color=colors[1], ax=ax1, alpha=0.25)
averagetimeseries(ub_S, color=colors[2], ax=ax1, alpha=0.25)
averagetimeseries(ub_U, color=colors[3], ax=ax1, alpha=0.25)
ax1.set_title('average time taken across space')
namemap = lesionordermap(info)
info = info.query(nonjumpers)
info = info.query("subject != 'JPAK_20'")
# Select data
group = list(names(info))
controls = list(names(control(info)))
lesions = list(names(lesion(info)))
matched = list(cagemates(lesion(info)))
selection = str.format("subject in {0}", group)
random = '(session == 13 and trial > 20) or (14 <= session < 17)'
steps = steps.query(random).query(selection)
cr = cr.query(random).query(selection)
steps = joinstepactivity(steps, cr, cract)
steps = getballistictrials(steps)
steps.xhead = flipleftwards(steps.xhead, steps.side)
normalize(steps, mediannorm, ['xhead', 'yhead'], level=['subject'])
steps = steps.rename(columns={'index': 'time'})
# Normalize data
cract.reset_index(inplace=True)
cract.set_index(['subject', 'session', 'crossing'], inplace=True)
cract = cract.join(steps, how='inner', rsuffix='R')
cract.xhead = flipleftwards(cract.xhead, cract.side)
cract.xhead_speed[cract.side == 'leftwards'] *= -1
bias = cract.eval('xheadR >= 0')
scract = cract[bias]
ucract = cract[~bias]
sb_SA = scract.query('stepstate3')
sb_UA = scract.query('not stepstate3')
ub_SA = ucract.query('stepstate3')