def go_particle():
img = Image()
detect = Detection()
found = []
initialState = None
pf = None
histograms = []
detections = []
hist_ref = None
#Recoleccion
while(True):
img.get()
#Face de recoleccion de información/ Aprendizaje
if len(histograms) < 10:
print len(histograms)
#Detect faces
found = detect.faces(img.image)
if len(found) > 0:
hist_ref = img.getColorHistogram(found[0]).ravel()
histograms.append(hist_ref)
img.show_hist(hist_ref)
hist_ref=hist_ref/float(hist_ref.sum())
else: #Aprendizaje finalizado
#Iniciar filtro, utilizar el ultimo ROI/found como initialState
# print len(hist_ref)
if initialState is None:
x,y,w,h = found[0]
hist = img.getColorHistogram(found[0]).ravel()
dx = random.uniform(0,5)
dy = random.uniform(0,5)
dw = random.uniform(0,.5)
dh = random.uniform(0,.5)
initialState = np.array([x,y,w,h,dx,dy,dw,dh])
ceroState = np.random.uniform(0,img.size[1],8)
cov = np.cov(np.asarray([initialState,ceroState]).T)
pf = ParticleFilter(initialState,cov,hist_ref,10.,100)
u=np.zeros((100,4))
for rect in pf.states:
rect = rect[:4]
img.draw_roi([rect])
# pf.update(img)
else: #ya se ha inicializado el filtro ahora se busca actualizar y predecir
old=pf.states
pf.predict(img.size[0],img.size[1],u)
pf.update(img)
for rect in pf.states:
rect = rect[:4]
img.draw_roi([rect])
u=(old-pf.states)[:,-4:]
img.show()
if 0xFF & cv2.waitKey(5) == 27:
break
def compare_distributions():
img = Image()
detect = Detection()
histograms = []
detections = []
while(True):
img.get()
#Detect faces
found = detect.faces(img.image)
if len(found) > 0:
# Set reference histogram from roi
# roi is the first face detected assuming one face in the portview
hist_ref = img.getColorHistogram(found[0]).ravel()
histograms.append(hist_ref)
detections.append(found)
img.show_hist(hist_ref)
hist_ref=hist_ref/float(hist_ref.sum())
if 0xFF & cv2.waitKey(5) == 27:
break
print np.asmatrix(histograms).shape
alpha_hat,it= fit_betabinom_minka_alternating(histograms)
gauss_data_f = []
gauss_data_n = []
polya_data_f = []
polya_data_n = []
while(True):
img.get()
found = detect.faces(img.image)
if len(found) > 0:
img.draw_roi([found[0]])
hist=img.getColorHistogram(found[0]).ravel()
img.show_hist(hist)
hist_norm=hist/float(hist.sum())
#For face
D=bhattacharyya(hist_ref,hist_norm)
G=np.exp(-20.*D**2)
P=np.exp(log_like_polya(alpha_hat,np.array([hist])))
#FOr not face
x = 0#random.sample([random.uniform(0,img.size[0]) for i in range(100)],1)[0]
y = 0#random.sample([random.uniform(0,img.size[1]) for i in range(100)],1)[0]
w = found[0][2]#random.sample([random.uniform(0,200) for i in range(100)],1)[0]
h = found[0][3]#random.sample([random.uniform(0,200) for i in range(100)],1)[0]
rect = (x,y,w,h)
hist = img.getColorHistogram(rect).ravel()
img.show_hist(hist,'NO FACE')
hist_norm=hist/float(hist.sum())
D2=bhattacharyya(hist_ref,hist_norm)
G2=np.exp(-20.*D2**2)
P2=np.exp(log_like_polya(alpha_hat,np.array([hist])))
print 'B={0}, B={1} | G={2}, G={3} | P={4}, P={5}'.format(D,D2,G,G2,P,P2)
gauss_data_n.append(G2)
gauss_data_f.append(G)
polya_data_n.append(P2)
polya_data_f.append(P)
img.show()
if 0xFF & cv2.waitKey(5) == 27:
cv2.destroyAllWindows()
break
gauss_data_f = np.asarray(gauss_data_f)
gauss_data_f = gauss_data_f[~np.isnan(gauss_data_f)]
gauss_data_n = np.asarray(gauss_data_n)
gauss_data_n = gauss_data_n[~np.isnan(gauss_data_n)]
d = np.concatenate((gauss_data_f,gauss_data_n))
v = np.concatenate(( np.ones(len(gauss_data_f)), np.zeros(len(gauss_data_n))))
fpr, tpr, thresholds = roc_curve(v, d)
roc_auc = auc(fpr, tpr)
print("Area under the ROC curve : %f" % roc_auc)
polya_data_f = np.asarray(polya_data_f)
polya_data_f = polya_data_f[~np.isnan(polya_data_f)] #Remove Nan elements
polya_data_n = np.asarray(polya_data_n)
polya_data_n = polya_data_n[~np.isnan(polya_data_n)] #Remove Nan elements
d = np.concatenate((polya_data_f,polya_data_n))
v = np.concatenate(( np.ones(len(polya_data_f)), np.zeros(len(polya_data_n))))
fpr2, tpr2, thresholds = roc_curve(v, d)
roc_auc2 = auc(fpr2, tpr2)
print("Area under the ROC curve : %f" % roc_auc2)
pl.subplot(212)
pl.plot(fpr, tpr, label='ROC curve (area = %0.2f)' % roc_auc)
pl.plot([0, 1], [0, 1], 'k--')
pl.xlim([0.0, 1.0])
pl.ylim([0.0, 1.0])
pl.xlabel('False Positive Rate')
pl.ylabel('True Positive Rate')
pl.title('ROC for Gaussian')
pl.subplot(211)
pl.plot(fpr2, tpr2, label='ROC curve (area = %0.2f)' % roc_auc2)
pl.plot([0, 1], [0, 1], 'k--')
pl.xlim([0.0, 1.0])
pl.ylim([0.0, 1.0])
pl.xlabel('False Positive Rate')
pl.ylabel('True Positive Rate')
pl.title('ROC for Polya')
pl.show()
