def whenSelected():
if getSelection():
if pcol.get() or p3d.get():
if (checkValidInt(sf, 0) and checkValidInt(ws, 1)
and checkValidInt(ch, 2) and checkBandPass(hi, lo)):
for file in files:
freqs, time, data = stft.main(params, file)
# print(data.shape)
if pcol.get():
plotting.plotPColor(data, file, params[2], freqs, time)
if p3d.get():
plotting.plot3D(data, file, params[2], freqs, time)
else:
mbox.showerror("No graph type selected", errors[-1])
# if(constants.URL.__contains__('sonar')) :
# best_fitness = output(best_fitness)
# print("BEST FITNESS ::",output(best_fitness))
print("BEST FITNESS ::", (best_fitness))
print("labelled :: \n", labelled)
##############################
#### FMEASURE
if (constants.URL.__contains__('ionosphere')
or constants.URL.__contains__('iris')
or constants.URL.__contains__('wine')
or constants.URL.__contains__('parkinsons')
or constants.URL.__contains__('sonar')
or constants.URL.__contains__('segmentation')
or constants.URL.__contains__('glass')):
# classdata = pd.read_csv(constants.cURL)
classdata = np.array(y_dataset_glob)
#print(classdata)
fmeasure(labelled, classdata)
##############################
if constants.URL.__contains__('random2'):
plot2D(got_back[best_index], pop_distri)
if constants.URL.__contains__('random3'):
plot3D(got_back[best_index], pop_distri)
bkgTauMin, bkgAMin, sortedTimes,
meanSigma) #Calculating pdf function with background
area2 = trapz(pdf2,
sortedTimes) #Calculating area of new pdf using trapezium rule
#Calculating errors
TauAErrors = minim.bkgError(bkgTaus, bkgAs, bkgTauMin, bkgAMin, bkgNllMin,
funcs.bkgNll)
meanTauError = np.mean([TauAErrors[0], TauAErrors[1]])
meanAError = np.mean([TauAErrors[2], TauAErrors[3]])
### CREATING AND DISPLAYING PLOTS ###
pt.plotContour(bkgTaus,
bkgAs,
funcs.bkgNll,
levels=np.arange(bkgNllMin + 0.5, bkgNllMin + 0.5 + 100, 10))
pt.plot3D(bkgTaus, bkgAs, funcs.bkgNll)
intersect, m, c = pt.plotErrorsVSReadings(sizes, sizeErrors)
pt.plotNLL(taus, likelihoods, tauMin, nllMin)
pt.plotHist(times, bins=100, sortedTimes=sortedTimes, pdfs=[pdf, pdf2])
### PRINTING RESULTS ###
print "Area under initial PDF (without background):", area
print "Area under refined PDF (with background):", area2
print "NLL at Minimum 1D:", nllMin
print "Tau at Minimum 1D:", tauMin, "picoseconds"
print "Positive Tau Error 1D:", posError, "picoseconds"
print "Negative Tau Error 1D:", negError, "picoseconds"
print "Mean Tau Error 1D:", meanError, "picoseconds"
print "Last Parabolic Estimate Error of Tau:", parabError, "picoseconds"
print "Intersect of extrapolation of gradient", m, "and intercept", c, "gives approximately", int(
10**intersect[0]), "readings required for accuracy of 0.001ps"
def test_plot3d(self):
data = dict([(i, dict([(j, random.random()) for j in range(10)]))
for i in range(5)])
plot3D(data)
plt.show()
labelled = fitness.get_pop_distri_label(file, centroid, constants.K)
print("labelled :: \n", labelled)
##############################
#### FMEASURE
if (constants.URL.__contains__('ionosphere')
or constants.URL.__contains__('iris')
or constants.URL.__contains__('wine')
or constants.URL.__contains__('parkinsons')
or constants.URL.__contains__('sonar')
or constants.URL.__contains__('segmentation')
or constants.URL.__contains__('glass')):
# classdata = pd.read_csv(constants.cURL)
classdata = np.array(y_dataset)
#print(classdata)
fmeasure(labelled, classdata)
##############################
if constants.URL.__contains__('random2'):
plot2D(centroid, pop_distri)
if constants.URL.__contains__('random3'):
plot3D(centroid, pop_distri)
#data = np.array(data)
#print(dist(centroids[0],data[0]),dist(centroids[1],data[0]),dist(centroids[2],data[0]))
#print(dist(centroids[0],data[55]),dist(centroids[1],data[55]),dist(centroids[2],data[55]))
#print(dist(centroids[0],data[127]),dist(centroids[1],data[127]),dist(centroids[2],data[127]))
def run():
global pts3
global pts4
if simulation and view:
plot.plot3D(data3D, 'Original 3D Data')
if view:
plot.plot2D(pts1_raw, name='First Statics (Noise not shown)')
plot.plot2D(pts2_raw, name='Second Statics (Noise not shown)')
# FUNDAMENTAL MATRIX
F = getFundamentalMatrix(pts1, pts2)
# ESSENTIAL MATRIX (HZ 9.12)
E, w, u, vt = getEssentialMatrix(F, K1, K2)
# PROJECTION/CAMERA MATRICES from E (HZ 9.6.2)
P1, P2 = getNormalisedPMatrices(u, vt)
P1_mat = np.mat(P1)
P2_mat = np.mat(P2)
# FULL PROJECTION MATRICES (with K) P = K[Rt]
KP1 = K1 * P1_mat
KP2 = K2 * P2_mat
print "\n> KP1:\n", KP1
print "\n> KP2:\n", KP2
# SYNCHRONISATION + CORRECTION
if rec_data and simulation is False:
print "---Synchronisation---"
pts3, pts4 = synchroniseGeometric(pts3, pts4, F)
pts3 = pts3.reshape((1, -1, 2))
pts4 = pts4.reshape((1, -1, 2))
newPoints3, newPoints4 = cv2.correctMatches(F, pts3, pts4)
pts3 = newPoints3.reshape((-1, 2))
pts4 = newPoints4.reshape((-1, 2))
elif simulation:
print "> Simulation: Use whole point set for reconstruction"
pts3 = pts1
pts4 = pts2
# Triangulate the trajectory
p3d = triangulateCV(KP1, KP2, pts3, pts4)
# Triangulate goalposts
if simulation is False:
goalPosts = triangulateCV(KP1, KP2, postPts1, postPts2)
# SCALING AND PLOTTING
if simulation:
if view:
plot.plot3D(p3d, 'Simulation Reconstruction')
reprojectionError(K1, P1_mat, K2, P2_mat, pts3, pts4, p3d)
p3d = simScale(p3d)
if view:
plot.plot3D(p3d, 'Scaled Simulation Reconstruction')
else:
# add the post point data into the reconstruction for context
if len(postPts1) == 4:
print "> Concatenate goal posts to trajectory"
pts3_gp = np.concatenate((postPts1, pts3), axis=0)
pts4_gp = np.concatenate((postPts2, pts4), axis=0)
p3d_gp = np.concatenate((goalPosts, p3d), axis=0)
scale = getScale(goalPosts)
scaled_gp_only = [[a * scale for a in inner] for inner in goalPosts]
scaled_gp = [[a * scale for a in inner] for inner in p3d_gp]
scaled = [[a * scale for a in inner] for inner in p3d]
if view:
plot.plot3D(scaled_gp, 'Scaled 3D Reconstruction')
reprojectionError(K1, P1_mat, K2, P2_mat, pts3_gp, pts4_gp, p3d_gp)
getMetrics(scaled, scaled_gp_only)
scaled_gp = transform(scaled_gp)
if view:
plot.plot3D(scaled_gp, 'Final (Reorientated) 3D Reconstruction')
if ground_truth_provided:
reconstructionError(data3D, scaled_gp)
# write X Y Z to file
outfile = open('sessions/' + clip + '/3d_out.txt', 'w')
for p in scaled_gp:
p0 = round(p[0], 2)
p1 = round(p[1], 2)
p2 = round(p[2], 2)
string = str(p0) + ' ' + str(p1) + ' ' + str(p2)
outfile.write(string + '\n')
outfile.close()
def main():
path = sys.argv[1]
folder = os.path.dirname(path)
data_3d = getData(path)
data_3d = np.array(data_3d, dtype='float32')
# Artificial cameras - both have the same intrinsics
K = tools.CalibArray(1000, 640, -360)
dist = (0, 0, 0, 0)
# Some numbers we use a lot
rt2 = math.sqrt(2)
rt2on2 = rt2 / 2
# Lots of rotation matrices...
# naming: z45cc = 45deg counter-clockwise rotation about z
z45cc = np.mat([[1 / rt2, -1 / rt2, 0], [1 / rt2, 1 / rt2, 0], [0, 0, 1]],
dtype='float32')
nothing = np.mat([[1, 0, 0], [0, 1, 0], [0, 0, 1]], dtype='float32')
z90cc = np.mat([[0, -1, 0], [1, 0, 0], [0, 0, 1]], dtype='float32')
z90cw = np.mat([[0, 1, 0], [-1, 0, 0], [0, 0, 1]], dtype='float32')
z180 = np.mat([[-1, 0, 0], [0, -1, 0], [0, 0, 1]], dtype='float32')
x180 = np.mat([[1, 0, 0], [0, -1, 0], [0, 0, -1]], dtype='float32')
y180 = np.mat([[-1, 0, 0], [0, 1, 0], [0, 0, -1]], dtype='float32')
x90cw = np.mat([[1, 0, 0], [0, 0, 1], [0, -1, 0]], dtype='float32')
x90cc = np.mat([[1, 0, 0], [0, 0, -1], [0, 1, 0]], dtype='float32')
y90cc = np.mat([[0, 0, 1], [0, 1, 0], [-1, 0, 0]], dtype='float32')
y90cw = np.mat([[0, 0, -1], [0, 1, 0], [1, 0, 0]], dtype='float32')
z90cc_vec, jacobian = cv2.Rodrigues(z90cc)
z90cc_vec, jacobian = cv2.Rodrigues(z90cc)
y90cc_vec, jacobian = cv2.Rodrigues(y90cc)
x90cc_vec, jacobian = cv2.Rodrigues(x90cc)
y45cc = np.mat([[rt2on2, 0, rt2on2], [0, 1, 0], [-rt2on2, 0, rt2on2]],
dtype='float32')
y45cw = np.mat([[rt2on2, 0, -rt2on2], [0, 1, 0], [rt2on2, 0, rt2on2]],
dtype='float32')
# cameras have different poses
tvec1 = (-5, 1, 14)
tvec2 = (5, 1, 13)
# Project the model into each camera
img_pts1 = project(data_3d, K, y45cc, tvec1)
img_pts2 = project(data_3d, K, y45cw, tvec2)
img_pts1 = np.reshape(img_pts1, (len(img_pts1), 2, 1))
img_pts2 = np.reshape(img_pts2, (len(img_pts2), 2, 1))
# Plot the model and images
plot.plot3D(data_3d)
plotImagePoints(img_pts1)
plotImagePoints(img_pts2)
# Add noise to the image data if desired
try:
noise = sys.argv[2]
img_pts1 = addNoise(float(noise), img_pts1)
img_pts2 = addNoise(float(noise), img_pts2)
except indexError:
continue
writeData(folder, img_pts1, img_pts2)
def test_plot3d(self):
data = dict([(i, dict([(j, random.random())for j in range(10)])) for i in range(5)])
plot3D(data)
plt.show()
def main():
path = sys.argv[1]
folder = os.path.dirname(path)
data_3d = getData(path)
data_3d = np.array(data_3d, dtype='float32')
# Artificial cameras - both have the same intrinsics
K = tools.CalibArray(1000, 640, -360)
dist = (0, 0, 0, 0)
# Some numbers we use a lot
rt2 = math.sqrt(2)
rt2on2 = rt2 / 2
# Lots of rotation matrices...
# naming: z45cc = 45deg counter-clockwise rotation about z
z45cc = np.mat([[1 / rt2, -1 / rt2, 0],
[1 / rt2, 1 / rt2, 0],
[0, 0, 1]], dtype='float32')
nothing = np.mat([[1, 0, 0], [0, 1, 0], [0, 0, 1]], dtype='float32')
z90cc = np.mat([[0, -1, 0], [1, 0, 0], [0, 0, 1]], dtype='float32')
z90cw = np.mat([[0, 1, 0], [-1, 0, 0], [0, 0, 1]], dtype='float32')
z180 = np.mat([[-1, 0, 0], [0, -1, 0], [0, 0, 1]], dtype='float32')
x180 = np.mat([[1, 0, 0], [0, -1, 0], [0, 0, -1]], dtype='float32')
y180 = np.mat([[-1, 0, 0], [0, 1, 0], [0, 0, -1]], dtype='float32')
x90cw = np.mat([[1, 0, 0], [0, 0, 1], [0, -1, 0]], dtype='float32')
x90cc = np.mat([[1, 0, 0], [0, 0, -1], [0, 1, 0]], dtype='float32')
y90cc = np.mat([[0, 0, 1], [0, 1, 0], [-1, 0, 0]], dtype='float32')
y90cw = np.mat([[0, 0, -1], [0, 1, 0], [1, 0, 0]], dtype='float32')
z90cc_vec, jacobian = cv2.Rodrigues(z90cc)
z90cc_vec, jacobian = cv2.Rodrigues(z90cc)
y90cc_vec, jacobian = cv2.Rodrigues(y90cc)
x90cc_vec, jacobian = cv2.Rodrigues(x90cc)
y45cc = np.mat([[rt2on2, 0, rt2on2],
[0, 1, 0],
[-rt2on2, 0, rt2on2]], dtype='float32')
y45cw = np.mat([[rt2on2, 0, -rt2on2],
[0, 1, 0],
[rt2on2, 0, rt2on2]], dtype='float32')
# cameras have different poses
tvec1 = (-5, 1, 14)
tvec2 = (5, 1, 13)
# Project the model into each camera
img_pts1 = project(data_3d, K, y45cc, tvec1)
img_pts2 = project(data_3d, K, y45cw, tvec2)
img_pts1 = np.reshape(img_pts1, (len(img_pts1), 2, 1))
img_pts2 = np.reshape(img_pts2, (len(img_pts2), 2, 1))
# Plot the model and images
plot.plot3D(data_3d)
plotImagePoints(img_pts1)
plotImagePoints(img_pts2)
# Add noise to the image data if desired
try:
noise = sys.argv[2]
img_pts1 = addNoise(float(noise), img_pts1)
img_pts2 = addNoise(float(noise), img_pts2)
except indexError:
continue
writeData(folder, img_pts1, img_pts2)
def run():
global pts3
global pts4
if simulation and view:
plot.plot3D(data3D, 'Original 3D Data')
if view:
plot.plot2D(pts1_raw, name='First Statics (Noise not shown)')
plot.plot2D(pts2_raw, name='Second Statics (Noise not shown)')
# FUNDAMENTAL MATRIX
F = getFundamentalMatrix(pts1, pts2)
# ESSENTIAL MATRIX (HZ 9.12)
E, w, u, vt = getEssentialMatrix(F, K1, K2)
# PROJECTION/CAMERA MATRICES from E (HZ 9.6.2)
P1, P2 = getNormalisedPMatrices(u, vt)
P1_mat = np.mat(P1)
P2_mat = np.mat(P2)
# FULL PROJECTION MATRICES (with K) P = K[Rt]
KP1 = K1 * P1_mat
KP2 = K2 * P2_mat
print "\n> KP1:\n", KP1
print "\n> KP2:\n", KP2
# SYNCHRONISATION + CORRECTION
if rec_data and simulation is False:
print "---Synchronisation---"
pts3, pts4 = synchroniseGeometric(pts3, pts4, F)
pts3 = pts3.reshape((1, -1, 2))
pts4 = pts4.reshape((1, -1, 2))
newPoints3, newPoints4 = cv2.correctMatches(F, pts3, pts4)
pts3 = newPoints3.reshape((-1, 2))
pts4 = newPoints4.reshape((-1, 2))
elif simulation:
print "> Simulation: Use whole point set for reconstruction"
pts3 = pts1
pts4 = pts2
# Triangulate the trajectory
p3d = triangulateCV(KP1, KP2, pts3, pts4)
# Triangulate goalposts
if simulation is False:
goalPosts = triangulateCV(KP1, KP2, postPts1, postPts2)
# SCALING AND PLOTTING
if simulation:
if view:
plot.plot3D(p3d, 'Simulation Reconstruction')
reprojectionError(K1, P1_mat, K2, P2_mat, pts3, pts4, p3d)
p3d = simScale(p3d)
if view:
plot.plot3D(p3d, 'Scaled Simulation Reconstruction')
else:
# add the post point data into the reconstruction for context
if len(postPts1) == 4:
print "> Concatenate goal posts to trajectory"
pts3_gp = np.concatenate((postPts1, pts3), axis=0)
pts4_gp = np.concatenate((postPts2, pts4), axis=0)
p3d_gp = np.concatenate((goalPosts, p3d), axis=0)
scale = getScale(goalPosts)
scaled_gp_only = [[a * scale for a in inner] for inner in goalPosts]
scaled_gp = [[a * scale for a in inner] for inner in p3d_gp]
scaled = [[a * scale for a in inner] for inner in p3d]
if view:
plot.plot3D(scaled_gp, 'Scaled 3D Reconstruction')
reprojectionError(K1, P1_mat, K2, P2_mat, pts3_gp, pts4_gp, p3d_gp)
getMetrics(scaled, scaled_gp_only)
scaled_gp = transform(scaled_gp)
if view:
plot.plot3D(scaled_gp, 'Final (Reorientated) 3D Reconstruction')
if ground_truth_provided:
reconstructionError(data3D, scaled_gp)
# write X Y Z to file
outfile = open('sessions/' + clip + '/3d_out.txt', 'w')
for p in scaled_gp:
p0 = round(p[0], 2)
p1 = round(p[1], 2)
p2 = round(p[2], 2)
string = str(p0) + ' ' + str(p1) + ' ' + str(p2)
outfile.write(string + '\n')
outfile.close()