Bernard Kwok

That's a good question. I did a small experiment and OIIO is also about 3x slower for performing the compare at least. About 7 seconds so better than PIL +...

@jstone-lucasfilm, @lgritz - I can replace the OpenCV code with OIIO and it can be reviewed here. - As for tests, I can place a sample here of a few...

Here are 2 pairs. The subset has 121 pairs to compare. [standard_surface_chess_set.zip](https://github.com/user-attachments/files/18769684/standard_surface_chess_set.zip)

Replace OpenCV with OIIO. Leaving OpenCV implementation for comparison / posterity ```python def computeDiff(image1Path, image2Path, imageDiffPath): try: # Remove the diff image if it already exists if os.path.exists(imageDiffPath): os.remove(imageDiffPath) #...

For comparison, as @jstone-lucasfilm notes it's not a huge difference so we can stick with OIIO to reduce non-ASWF dependencies as suggested and incorporate your changes @lgritz. (@fpliu I didn't...

I very much like this approach. Some basic comments which I'm not sure make sense since you seem to have already worked out a lot of details which I don't...

Sample script which will behave different if `importLibrary` is used instead of `setDataLibrary`. ```python import MaterialX as mx doc = mx.createDocument() #graph = doc.addNodeGraph("graph1") stdlib = mx.createDocument() mx.loadLibraries(mx.getDefaultDataLibraryFolders(), mx.getDefaultDataSearchPath(), stdlib)...

I tried creating a virtual environment and installed only MaterialX 1.39.4 using pip and ran this script with Python 3.11.0. Other than taking a while to run I didn't get...

Checking with `nanogui` (Viewer) and `imgui` (Editor) both appear to assume sRGB for display colors so looks like the reverse transform is required.