question

[Llj-l]

Dear Dr. Nie! I have been reproducing your work recently and want to further apply it to complete the stitching of the target image and the reference image, so as to achieve the effect of a complete stitching, but I don't know how to do it, could you please give me an answer? Thank you.

[nie-lang]

It's hard to be used in image stitching.
If you read Section ⅢC of the paper, you can find the mesh warp used in our paper differs from the traditional mesh warp. This is because we have to implement the parallel computation in a deep learning framework, thus we modify the traditional mesh warp to the "backward" mesh warp.
If you want to use the mesh warp in deep image stitching, I have 3 suggestions:

1. implement the "forward" mesh warp process in a parallel manner (it may be difficult)
2. use the nearest homography(the homo that the nearest grid corresponds) for the non-overlap regions
3. warp the images (in the stitching-domain) using a global homography first, then apply the deep mesh warp on the warped images (it may work, but the grids should be dense in my view)