Images getting generated differently about imagetransformations.jl HOT 4 CLOSED

I don't quite understand the meaning of the results, but since you figured it out I'm going to close it.

from imagetransformations.jl.

• Instead of h=homography2d(srcs,tgts), can you start with hardcoded transformation matrix? Because no tests are written for homography2d, I'm skeptical of this statement "using the same homography matrix".
• Instead of using big real images like this, can you try it on very small images (e.g., reshape(collect(0:0.04:0.96), 5, 5))?

from imagetransformations.jl.

• I used homography2d to get the homography matrix, and then hardcoded that matrix to Kornia which gave me the above results. Therefore, I don't think homography2d is giving us a problem.
• I will try using smaller images as you have suggested.

from imagetransformations.jl.

I have gotten why this was going south.
Plot.jl is plotting according to the transposed dimensions. Surprisingly, FileIO was also reading the images according to transposed dimensions.
[i, j] in images/array indexing mean that i is along the vertical axis and j is along the horizontal axis.
In Plots.jl, it plots the i along the horizontal axis and j along the vertical axis. Hence the issue.
With small changes, like this -

src_img = load("source.jpeg")'
tgt_img = load("target.jpeg")'
...
plot(mosaicview(tgt_img,warp_me2(src_img,h);nrow=1))

This is the correct warp I get, of course transposed image -

from imagetransformations.jl.