def main():
# The model
model_path = os.path.join(util.get_chira_root(), "data/models/hand-model-v3/exported_template_from_blender")
model = load_model(model_path)
viewer = ModelViewer(model, optimizer.Optimizer(model_path, model))
viewer.sequence = local_annotation_sequence()
viewer.start()
def main():
# The model
model_path = os.path.join(
util.get_chira_root(),
'data/models/hand-model-v3/exported_template_from_blender')
model = load_model(model_path)
viewer = ModelViewer(model, optimizer.Optimizer(model_path, model))
viewer.sequence = local_annotation_sequence()
viewer.start()
import os
from pychira.model import load_model
from pychira import util
util.add_ezvtk_to_path()
import ezvtk.vis
import ezvtk.troupe
import vtk
model_path = os.path.join(
util.get_chira_root(),
'data/models/hand-model-v3/exported_template_from_blender')
model = load_model(model_path)
class ModelViewer():
def __init__(self):
self.viewer = ezvtk.vis.Viewer()
# Add triangular mesh.
self.mesh_troupe = ezvtk.troupe.MeshTroupe()
self.mesh_troupe.set_polygons(model.triangles)
self.mesh_troupe.set_points(model.base_positions)
# Also add points for easy picking with the mouse.
self.vertices_troupe = ezvtk.troupe.SpheresTroupe(color=(.2, .5, .7),
radius=0.002)
#for i_data_point in range(n_data_points):
# energy.data_points[i_data_point] = vertex_positions[correspondences[i_data_point]]
#print('Energy with perfect data point correspondences:', energy.evaluate(theta))
## Now add some noise to the data points.
#energy.data_points += .1 * np.random.randn(*energy.data_points.shape)
##save_instance("C:/Users/t-filsra/Workspace/autodiff/hand/instance.txt", correspondences, data_points, theta);
#print('Energy with noisy data points:', energy.evaluate(theta))
#### generate instances ####
n_data_points = [100, 192, 200, 400, 800, 1600, 3200, 6400, 12800, 25600, 51200, 100000]
data_dir = "C:/Users/t-filsra/Workspace/autodiff/hand_instances/complicated_big/"
model = load_model(data_dir + "model/")
for i in range(len(n_data_points)):
### complicated version ###
correspondences = np.random.random_integers(0, model.triangles.shape[0] - 1, n_data_points[i])
data_points = np.zeros((n_data_points[i], 3))
energy = Energy(model, data_points, correspondences)
theta = np.random.randn(energy.n_theta)
us = np.random.rand(n_data_points[i], 2)*0.5;
pose_params = energy.to_pose_params(theta)
vertex_positions = model.get_skinned_vertex_positions(pose_params)
for i_data_point,u in zip(range(n_data_points[i]),us):
"""A script to display the hand model."""
import os
from pychira.model import load_model
from pychira import util
util.add_ezvtk_to_path()
import ezvtk.vis
import ezvtk.troupe
import vtk
model_path = os.path.join(util.get_chira_root(), 'data/models/hand-model-v3/exported_template_from_blender')
model = load_model(model_path)
class ModelViewer():
def __init__(self):
self.viewer = ezvtk.vis.Viewer()
# Add triangular mesh.
self.mesh_troupe = ezvtk.troupe.MeshTroupe()
self.mesh_troupe.set_polygons(model.triangles)
self.mesh_troupe.set_points(model.base_positions)
# Also add points for easy picking with the mouse.
self.vertices_troupe = ezvtk.troupe.SpheresTroupe(color=(.2, .5, .7), radius=0.002)
self.vertices_troupe.set_positions(model.base_positions)
# energy.data_points[i_data_point] = vertex_positions[correspondences[i_data_point]]
#print('Energy with perfect data point correspondences:', energy.evaluate(theta))
## Now add some noise to the data points.
#energy.data_points += .1 * np.random.randn(*energy.data_points.shape)
##save_instance("C:/Users/t-filsra/Workspace/autodiff/hand/instance.txt", correspondences, data_points, theta);
#print('Energy with noisy data points:', energy.evaluate(theta))
#### generate instances ####
n_data_points = [
100, 192, 200, 400, 800, 1600, 3200, 6400, 12800, 25600, 51200, 100000
]
data_dir = "C:/Users/t-filsra/Workspace/autodiff/hand_instances/complicated_big/"
model = load_model(data_dir + "model/")
for i in range(len(n_data_points)):
### complicated version ###
correspondences = np.random.random_integers(0,
model.triangles.shape[0] - 1,
n_data_points[i])
data_points = np.zeros((n_data_points[i], 3))
energy = Energy(model, data_points, correspondences)
theta = np.random.randn(energy.n_theta)
us = np.random.rand(n_data_points[i], 2) * 0.5
pose_params = energy.to_pose_params(theta)