Numpy array to mask/image?

[tkwasnitschka]

I want to create masks from depth maps using numpy. Haven't checked all my code yet but first of all, how do I convert my numpy array back to a photoscan mask or image?

Code: [Select]

import PhotoScan, numpy

chunk = PhotoScan.app.document.chunk
scale = chunk.transform.scale
camera = chunk.cameras[0]
depth = chunk.model.renderDepth(camera.transform, camera.sensor.calibration) #unscaled depth
threshold = 4

# convert to numpy array:
map = numpy.frombuffer(depth.tostring(), dtype=numpy.float32)

# scale array:
map_scaled = map * scale

# apply threshold:
mask = ((map > threshold) * 255).astype("uint8") # <-- is this actually right...?

# write back to image:
camera.mask.image() = mask    # this is not right...


sorry for the basic question, many thanks!
Tom

[Alexey Pasumansky]

Hello Tom,

At first you need to create an image (PhotoScan.Image) and then pass it for mask:
mask = PhotoScan.Mask()
mask.setImage(image)

In the following thread there's a post regarding image object creation from numpy array:
http://www.agisoft.com/forum/index.php?topic=8225.msg39597#msg39597

[tkwasnitschka]

Thanks Alexey,
yes I have been staring at this post all afternoon, but I need the other way around: write the contents of an array back to the photoscan image. That post only talks about extracting an image as a numpy array, right?
If I follow your instructions, I get:

Code: [Select]

PhotoScan.Image = mask

mask
Out[60]: 2018-08-29 16:32:52 array([0, 0, 0, ..., 0, 0, 0], dtype=uint8)

PhotoScan.Image
Out[61]: 2018-08-29 16:33:01 array([0, 0, 0, ..., 0, 0, 0], dtype=uint8)

mask = PhotoScan.Mask()

mask
Out[63]: 2018-08-29 16:33:43 <PhotoScan.Mask at 0x1dfc7af8>

mask.setImage(PhotoScan.Image)
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-65-8fdb8ae6c111> in <module>()
----> 1 mask.setImage(PhotoScan.Image)

TypeError: argument 1 must be PhotoScan.Image, not numpy.ndarray

cheers
Tom

[tkwasnitschka]

Alexey, I have seen that post but I dont get it.
Please, how do I convert a numpy.ndarray back to a photoscan image?
Thanks so much
Tom

[tkwasnitschka]

This is my updated script. It creates a mask contained in the alpha channel but I fail to load it back to Photoscan.
The mask always returns the full image area and disregards what I saved in the A channel. Saving the Photoscan image, I see the mask is contained in the image. What am I doing wrong?

Code: [Select]

import PhotoScan, cv2
import numpy as np

print("start")
chunk = PhotoScan.app.document.chunk
scale = chunk.transform.scale
camera = chunk.cameras[0]
image = camera.photo.image()
depth = chunk.model.renderDepth(camera.transform, camera.sensor.calibration) #unscaled depth
map = np.fromstring(depth.tostring(), dtype=np.float32)

# scale array:
map_scaled = map*scale

# apply treshold:
threshold = 4
mask = ((map_scaled > threshold) * 255).astype("uint8")

# write back:
mask_img = PhotoScan.Image.fromstring(mask, image.width, image.height, 'K', datatype='U8')  <-- WRONG??
camera.mask = PhotoScan.Mask()
camera.mask.setImage(mask_img)
UPDATE: Found the Error. The channel must be K, not A. Updated the code above.

[Alexey Pasumansky]

Hello Tom,

Can you just try:

Code: [Select]

mask_img = PhotoScan.Image.fromstring(mask, image.width, image.height, ' ', datatype='U8')