# Batch Processing

(batch_processing)=

```{note}

It's not faster than a loop with ``GLCMCross``
```

For multiple images of the **SAME SIZE**, you may run them concurrently.

This is slightly faster on ``GLCM`` as it will start on the next image while the current is still being processed.

```pycon
>>> from glcm_cupy import GLCM
>>> import cv2
>>> ar0 = cv2.imread("image_0.jpg")
>>> ar1 = cv2.imread("image_1.jpg")
>>> ar0.shape
(1080, 1920, 3)
>>> # This is why you need them to be the same size
>>> g = GLCM(np.stack([ar0, ar1])).run(ar)
>>> g.shape
(2, 1074, 1914, 3, 7)
>>> g0, g1 = g[0], g[1]
```

The dimensions here are ``(batch, height, width, channel, features)``.

So, to retrieve the 1st image's glcm, `g[0]`

## GLCMCross

```{note}

It's not faster than a loop with ``GLCMCross``,  however it may be simpler in syntax.
```

```pycon
>>> g = GLCMCross(np.stack([ar0, ar1])).run(ar)
>>> g.shape
(2, 1074, 1914, 3, 7)
>>> g0, g1 = g[0], g[1]
```

Behavior of the ``ix_combos`` specified does not change.

If it's blank, it'll generate all combinations for each image.