[GUI] Add handy ti.imscale(img, w, h) for scaling image #1735
Conversation
archibate
requested review from
yuanming-hu,
Rullec,
JYLeeLYJ and
taichi-gardener
Codecov Report
@@ Coverage Diff @@
## master #1735 +/- ##
==========================================
- Coverage 43.45% 43.35% -0.11%
==========================================
Files 44 44
Lines 6015 6053 +38
Branches 1042 1045 +3
==========================================
+ Hits 2614 2624 +10
- Misses 3250 3275 +25
- Partials 151 154 +3
Continue to review full report at Codecov.
|
docs/gui.rst
Outdated
| @@ -522,3 +522,14 @@ Image I/O | |||
| This function will create an instance of ``ti.GUI`` and show the input image on the screen. | |||
|
|
|||
| It has the same logic as ``ti.imwrite`` for different datatypes. | |||
|
|
|||
| .. function:: ti.imscale(img, w=512, h=None): | |||
There was a problem hiding this comment.
| .. function:: ti.imscale(img, w=512, h=None): | |
| .. function:: ti.imscale(img, w=512, h=512): |
I would suggest to use a meaningful number in the parameter h, because it can guide our users better even though when h is None it will be the same as w.
There was a problem hiding this comment.
How about ti.imscale(img, w=512, h=w)?
Co-authored-by: Xudong Feng <[email protected]>
| u, v = (img.shape[0] - 1) / (w - 1), (img.shape[1] - 1) / (h - 1) | ||
| x = np.clip(np.arange(w) * u, 0, img.shape[0] - 1).astype(np.int32) | ||
| y = np.clip(np.arange(h) * v, 0, img.shape[1] - 1).astype(np.int32) | ||
| return img[tuple(np.meshgrid(x, y))].swapaxes(0, 1) |
There was a problem hiding this comment.
Why is swapaxes necessary here?
There was a problem hiding this comment.
np.meshgrid is row-major (y, x), while our image is column major (x, y).
There was a problem hiding this comment.
OK sounds good. It's still very confusing to me, we should enforce correctness via a test.
There was a problem hiding this comment.
I can now verify them by clicking the four endpoint of window of game_of_life.py.
| u, v = (img.shape[0] - 1) / (w - 1), (img.shape[1] - 1) / (h - 1) | ||
| x = np.clip(np.arange(w) * u, 0, img.shape[0] - 1).astype(np.int32) | ||
| y = np.clip(np.arange(h) * v, 0, img.shape[1] - 1).astype(np.int32) | ||
| return img[tuple(np.meshgrid(x, y))].swapaxes(0, 1) |
There was a problem hiding this comment.
OK sounds good. It's still very confusing to me, we should enforce correctness via a test.
Co-authored-by: Yuanming Hu <[email protected]>
Related issue = #
[Click here for the format server]
When I was debugging MGPCG, I realized that a small resolution is fast but hardly-visible. So it would be great if I could scale that small image up.