def create_error_model(model, vectors=None, terror=0.0, rerror=0.0):
emodel = A.Model(task_params=model._task_params)
convert = {}
if not vectors:
vectors = {}
for camera in model.cameras:
emodel[camera] = A.Camera(camera, model[camera]._params,
pose=model[camera]._pose, mount=model[camera].mount)
if not camera in vectors:
vectors[camera] = (A.random_unit_vector(), A.random_unit_vector())
emodel[camera].set_absolute_pose(\
A.pose_error(model[camera].pose, vectors[camera][0], terror,
vectors[camera][1], rerror))
convert[camera] = -model[camera].pose + emodel[camera].pose
for sceneobj in model.keys():
if not sceneobj in model.cameras:
emodel[sceneobj] = model[sceneobj]
return emodel, convert, vectors
if experiment.exit:
break
current_frames += 1
normal = (f(t / float(opts.interpolate)) - \
f((t - 1) / float(opts.interpolate))).normal
angle = A.Point((0, -1, 0)).angle(normal)
axis = A.Point((0, -1, 0)) ** normal
R = A.Rotation.from_axis_angle(angle, axis)
experiment.model[targetobj].set_absolute_pose(\
A.Pose(T=f(t / float(opts.interpolate)), R=R))
if opts.visualize:
experiment.model[targetobj].update_visualization()
current = best
if opts.terror:
etarget.set_absolute_pose(A.pose_error(experiment.relevance_models[\
targetrm].pose, A.random_unit_vector(), opts.terror[0],
A.random_unit_vector(), opts.terror[1]))
elif opts.posetrack and emodel:
etarget.set_absolute_pose(experiment.relevance_models[targetrm].pose)
if current and score and opts.posetrack and emodel:
etarget.set_absolute_pose(etarget.pose + convert[current])
best, score = (emodel or experiment.model).best_view(etarget or \
experiment.relevance_models[targetrm], current=(current and \
frozenset([current]) or None), threshold=opts.threshold,
candidates=((vision_graph and score) and [frozenset(c) for c in \
vision_graph.neighbors(current) | set([current])] or None))
if emodel or etarget:
score = experiment.model.performance(\
experiment.relevance_models[targetrm], subset=best)
if not optcached:
optperf += experiment.model.best_view(\
