How to stack mutiple images numpy ndarray

Question

I get the image numpy array from opencv， then i want to stack 32 images together, the final shape i want to obtain is (3, 32, image_height, image_width), below is the snippet of code :

import cv2
import numpy as np
video_path = 'xxxx.mp4'
frame_buffer = np.array([])
frame_index = 0
frame_buffer_num = 0
cap = cv2.VideoCapture(video_path)
while True:
  ret, image_np = cap.read()
  print(image_np.shape)
  if frame_index == 0:
    frame_buffer = image_np  # initialize empty frame_buffer
    frame_index += 1
    frame_buffer_num += 1
    continue

  frame_index += 1
  frame_buffer_num += 1
  frame_buffer = np.stack(frame_buffer, image_np)
  if frame_buffer_num == 32:
      print(frame_buffer.shape)
      break

I run it but got following error:

Traceback (most recent call last):
  File "/home/user/Python/temp.py", line 19, in <module>
    frame_buffer = np.stack(frame_buffer, image_np)
  File "<__array_function__ internals>", line 6, in stack
  File "/home/user/miniconda3/lib/python3.7/site-packages/numpy/core/shape_base.py", line 430, in stack
    axis = normalize_axis_index(axis, result_ndim)
TypeError: only size-1 arrays can be converted to Python scalars

-----------UPDATE-------------

Thanks for @abe 's inspiration, The following code works :

import cv2
import numpy as np
video_path = 'xxxx.mp4'
frame_index = 0
frame_buffer_num = 0
cap = cv2.VideoCapture(video_path)
width = cap.get(cv2.CAP_PROP_FRAME_WIDTH)  # float
height = cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
frame_buffer = np.zeros((3, 32, int(height), int(width)))
while True:
  ret, image_np = cap.read()
  image_np = image_np.reshape(3, int(height), int(width))
  print(image_np.shape)
 
  frame_buffer[:, frame_buffer_num, :, :] = image_np
  frame_index += 1
  frame_buffer_num += 1

  if frame_buffer_num == 32:
      print(frame_buffer)
      print(frame_buffer.shape)
      break

abe · Accepted Answer · 2020-12-31 14:53:17Z

2

The image you get for each frame has a shape (3, H, W) right? And you want to stack 32 of these. Then, you can first reshape image_np as image_np = image_np.reshape((3, 1, H, W)), then np.append(frame_buffer, image_np, axis=1) where frame_buffer was initialized with the first reshaped instance of image_np. That should result in a tensor with shape (3, 32, H, W)

Alternatively, you can initialize frame_buffer = np.zeros((3, 32, H, W)) and at each iteration, frame_buffer[:, i, :, :] = image_np

answered Dec 31, 2020 at 14:53

abe

9897 silver badges10 bronze badges

Sign up to request clarification or add additional context in comments.

1 Comment

Wade Wang Over a year ago

Thank you for providing such ingenious and workable solution ! I have successfully run through it by using the alternative mthod. What I got from opencv is actually (H, W, 3), but this is not a big problem. I can add a line of code :image_np = image_np.reshape(3, H, W)

Collectives™ on Stack Overflow

How to stack mutiple images numpy ndarray

1 Answer 1

1 Comment

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

1 Comment

Your Answer

Sign up or log in

Post as a guest

Related