Capturing pose
In this example, we show how to capture the bounding box, object centre, and local object axes.

"""Script to show how to capture the pose of 3D objects."""
import blendersynth as bsyn
bsyn.run_this_script(blend_src='../resources/scenes/capturing_pose.blend') # load from scene
# Load Blender script and set-up render properties
bsyn.render.set_resolution(512, 512)
bsyn.render.set_cycles_samples(10)
# Get objects
monkey = bsyn.Mesh.from_scene('Monkey')
cube = bsyn.Mesh.from_scene('Cube')
# Define compositor and visuals
comp = bsyn.Compositor()
bounding_box_visual = comp.get_bounding_box_visual()
keypoints_visual = comp.get_keypoints_visual(marker='x') # Create a visual of keypoints
axis_visual = comp.get_axes_visual()
combined_visual = comp.stack_visuals(bounding_box_visual, keypoints_visual, axis_visual)
# Define outputs
comp.define_output('Image')
comp.define_output(combined_visual, name='visual')
# Calculate bounding boxes
objects = [monkey, cube]
bounding_boxes = bsyn.annotations.bounding_boxes(objects)
# Calculate object centres as keypoints
keypoints = bsyn.annotations.keypoints.project_keypoints([obj.centroid() for obj in objects])
# Calculate axes in 2D. In 3D, use e.g. monkey.axes
axes = bsyn.annotations.get_axes(objects)
render_result = comp.render(annotations=axes + keypoints + bounding_boxes)
render_result.save_all('pose')