11

How to display image in standard numpy/opencv/matplotlib format in Kivy? Kivy uses different image memory layout and I can't figure out which one.

The following code works totally fine. Image was captured using cv2 VideoCapture. I think image is BGR, array dimensions are (360, 480, 3):

    ret, image = video_capture.read()
    cv2.imshow('image', image)
    cv2.waitKey()

Trying to display it with the following code produce messy results:

    video_texture = Texture.create(size=image.shape[:2])
    video_texture.blit_buffer(image.tostring(), colorfmt='rgb', bufferfmt='ubyte')

    # ...

    video_panel = self.ids['video_panel']
    with video_panel.canvas:
        Rectangle(texture=video_texture, pos=video_panel.pos, size=video_panel.size)
DikobrAz
  • 3,557
  • 4
  • 35
  • 53

2 Answers2

4

Found the right transformation. Probably suboptimal:

    ret, image = video_capture.read()
    image = np.rot90(np.swapaxes(image, 0, 1))
    video_texture = Texture.create(size=(image.shape[1], image.shape[0]), colorfmt='rgb')
    video_texture.blit_buffer(image.tostring(), colorfmt='bgr', bufferfmt='ubyte')
DikobrAz
  • 3,557
  • 4
  • 35
  • 53
3

I found that the rotation can also be done using the numpy function flipud(), not sure if it is actually faster (I'm not that experienced with python/numpy) but for some this might be more logical to read instead of the 90 degrees rotation and the swapaxes.

I. Hoekstra
  • 41
  • 2