plt.ylabel(' MSE ')
s = 0
for curv_type in curv_types:
err= list()
for sigma in Sigma:
cc = 'hex_quad_k1_' + str(Ks[i][0]) + '_k2_' + str(Ks[i][1]) + '_' + str(nstep) + '_noise_' + str(
int(sigma * 1000))
# os.chdir(r"/hpc/meca/users/souani.a/curvature/Compute_curvatures/Noise")
# os.mkdir(cc )
# os.chdir(r"/hpc/meca/users/souani.a/curvature/Compute_curvatures/Noise/"+cc)
# sio.write_mesh(mesh, cc+'.gii')
list_exampels.append(cc)
mesh_file = 'hex_quad_k1_' + str(Ks[i][0]) + '_k2_' + str(Ks[i][1]) + '_' + str(nstep) + '.gii'
os.chdir("/hpc/meca/users/souani.a/curvature/Compute_curvatures/Curvatures/0Noise")
mesh = sio.load_mesh(mesh_file)
xx = sps.quadric_curv_mean(Ks[i][0], Ks[i][1])(mesh.vertices[:, 0], mesh.vertices[:, 1])
xx_m = np.zeros((nstep ** 2, 1))
texture_file = cc + '_curv_mean_' + curv_type + '.gii'
# mesh.apply_transform(mesh.principal_inertia_transform)
os.chdir(
"/hpc/meca/users/souani.a/curvature/Compute_curvatures/Curvatures/Noise/" + cc)
tex = sio.load_texture(texture_file)
err.append(mean_squared_error(xx, tex.darray))
plt.plot( Sigma,err, color=couleur[s] , marker= 'o', label=str(curv_types[s]))
print("Done for "+ cc[:-5]+ ' for '+ curv_type)
# plt.text(Sigma[0], err[0], curv_types[s], fontsize=9)
s += 1
# plt.show()
os.chdir(r"/hpc/meca/users/souani.a/curvature/Compute_curvatures/Curvatures/Figures/Mean_squarred_error_function_of_noise")
plt.savefig("MSE function of noise for " + cc)
for curv_type in curv_types:
mesh_file = 'hex_quad_k1_' + str(Ks[i][0]) + '_k2_' + str(Ks[i][1]) + '_' + str(nstep) + '.gii'
texture_file = cc + '_curv_mean_' + curv_type + '.gii'
os.chdir("/hpc/meca/users/souani.a/curvature/Compute_curvatures/Quadrics")
mesh = sio.load_mesh(mesh_file)
# mesh.apply_transform(mesh.principal_inertia_transform)
os.chdir(
"/hpc/meca/users/souani.a/curvature/Compute_curvatures/Noise/" + cc)
tex = sio.load_texture(texture_file)
# splt.pyglet_plot(mesh, tex.darray, plot_colormap=True)
# splt.pyglet_plot(mesh, tex.darray, 'hot', plot_colormap=True)
# print(tex.darray)
data.append(tex.darray)
xx = sps.quadric_curv_mean(Ks[i][0], Ks[i][1])(mesh.vertices[:, 0], mesh.vertices[:, 1])
xx = np.reshape(xx, (1, nstep ** 2))
data = np.array(data)
data_m = np.zeros((6, nstep ** 2))
data_m[:5, :] = data[:, :, 0]
data_m[5, :] = xx
data_m = np.array(data_m)
embedding = MDS(n_components=2)
data_t = embedding.fit_transform(data_m[:, :])
# print(data_t)
fig = plt.figure()
ax = fig.add_subplot(111)
fig.suptitle(cc)
plt.xlabel('First component')
mesh = trimesh.Trimesh(faces=faces, vertices=coords, process=False)
# Show the quadric
# mesh.show()
# Estimate curvatures by Rusinkiewicz method
PrincipalCurvatures, PrincipalDir1, PrincipalDir2 = CC.GetCurvaturesAndDerivatives(mesh)
gaussian_curv = PrincipalCurvatures[0, :] * PrincipalCurvatures[1, :]
mean_curv = 0.5 * (PrincipalCurvatures[0, :] + PrincipalCurvatures[1, :])
# Calculate mean error between real and estimated curvatures
#Mean curvature
mesh_coords = mesh.vertices
curv_mean_real = gqs.quadric_curv_mean(Ks[0][0], Ks[0][1])(mesh_coords[:, 0], mesh_coords[:, 1])
print("Error = ", 20*np.log(np.mean(np.abs(mean_curv / curv_mean_real))))
#Gauss curvature
curv_gauss_real = gqs.quadric_curv_gauss(Ks[0][0], Ks[0][1])(mesh_coords[:, 0], mesh_coords[:, 1])
print("Error = ", 20*np.log(np.mean(np.abs(gaussian_curv / curv_gauss_real))))
# Plot mean curvature
vect_col_map = \
trimesh.visual.color.interpolate(mean_curv, color_map='jet')
if mean_curv.shape[0] == mesh.vertices.shape[0]:
mesh.visual.vertex_colors = vect_col_map
# Generate the quadric
# X, Y, faces, Zs = gqs.generate_quadric([ks], nstep=list_nbr_steps[i], ax=list_ax_ay[i][0ay=list_ax_ay[i][1])
# Z = Zs[0]
# coords = np.array([X, Y, Z]).transpose()
# mesh = trimesh.Trimesh(faces=faces, vertices=coords, process=False)
# mesh.show()
# print("mesh surface = ", mesh.area, "\n surface needed = ", list_surfaces[i])
os.chdir(
r"/hpc/meca/users/souani.a/curvature/Compute_curvatures/Curvatures/0Noise"
)
cc = 'hex_quad_k1_' + str(ks[0]) + '_k2_' + str(
ks[1]) + '_' + str(list_nbr_steps[i])
mesh = sio.load_mesh(cc + '.gii')
data = list()
xx = gqs.quadric_curv_mean(ks[0], ks[1])(mesh.vertices[:, 0],
mesh.vertices[:, 1])
print("-------------------------------" + cc + '_noise_' +
str(int(sigma * 1000)))
for curv_type in curv_types:
print(curv_type)
texture_file = cc + '_noise_' + str(int(
sigma * 1000)) + '_curv_mean_' + curv_type + '.gii'
# mesh.apply_transform(mesh.principal_inertia_transform)
os.chdir(
"/hpc/meca/users/souani.a/curvature/Compute_curvatures/Curvatures/Noise/"
+ cc + '_noise_' + str(int(sigma * 1000)))
tex = sio.load_texture(texture_file)
data.append(tex.darray)
data = np.array(data)