3

I am a beginner in Blender. This is what I do to save the depth: enter image description here

I can get the real Z in meter from the camera to the object. But how to simulate the field of view of the depth sensor?

Concretely, this is what I would like to do but using some mathematical nodes:

if (depth(i,j) < z_max) {
  if (depth(i,j) > z_min) {
    depth_clean(i,j) = depth(i,j);
  } else {
    depth_clean(i,j) = 0;
  }
} else {
  depth_clean(i,j) = 0;
}
Catree
  • 131
  • 2

1 Answers1

0

This is how to do it using the Python API, see this answer for a reference:

  • add a viewer node and connect the Z ouput of the Render Layers to the Viewer node
  • get the pixel data of the viewer
  • perform the depth "clipping" in Python
  • copy back to a bpy.ops.image
  • save the image in OpenEXR format

Script:

import bpy
import numpy as np

# https://blender.stackexchange.com/questions/35191/how-to-get-color-and-z-depth-from-viewer-node/71264#71264
bpy.context.scene.render.use_compositing = True
bpy.context.scene.use_nodes = True
tree = bpy.context.scene.node_tree
links = tree.links

for n in tree.nodes:
    tree.nodes.remove(n)
rl = tree.nodes.new('CompositorNodeRLayers')

vl = tree.nodes.new('CompositorNodeViewer')
vl.use_alpha = True
links.new(rl.outputs[0], vl.inputs[0])  # link Render Image to Viewer Image
links.new(rl.outputs[2], vl.inputs[1])  # link Render Z to Viewer Alpha

#render
bpy.context.scene.render.resolution_percentage = 100 #make sure scene height and width are ok (edit)

step_count = 251
scene = bpy.context.scene
for step in range(1, step_count):
    # Set render frame
    scene.frame_set(step)

    # Render
    bpy.ops.render.render()

    #get the pixels and put them into a numpy array
    pixels = np.array(bpy.data.images['Viewer Node'].pixels)

    width = bpy.context.scene.render.resolution_x 
    height = bpy.context.scene.render.resolution_y

    #reshaping into image array 4 channel (rgbz)
    image = pixels.reshape(height,width,4)

    #get only depth
    z = image[:,:,3]

    #simulate depth sensor fov / depth range
    def clip_depth(z, z_near, z_far, default_value=0):
        if z > z_far:
            return default_value
        elif z < z_near:
            return default_value
        else:
            return z

    clip_func = np.vectorize(clip_depth, otypes=[np.float])
    z_near = 1
    z_far = 8
    z_clip = clip_func(z, z_near, z_far)

    #save depth to OpenEXR format
    image_settings = bpy.context.scene.render.image_settings
    image_settings.file_format = "OPEN_EXR"
    image_settings.color_depth = '32'

    #create an HDR image with name 'img_z'
    image_name = 'img_z'
    bpy.ops.image.new(name=image_name, width=z_clip.shape[1], height=z_clip.shape[0], alpha=True, float=True)
    img_z_clip = bpy.data.images[image_name]

    z_clip_rgba = np.dstack((z_clip, z_clip, z_clip, z_clip))
    img_z_clip.pixels = z_clip_rgba.ravel()
    img_z_clip.save_render('/tmp/blender_z_depth_clip_%04d.exr' % step)
Catree
  • 131
  • 2