search

lhaippp / RecDiffusion

[CVPR2024]: RecDiffusion: Rectangling for Image Stitching with Diffusion Models
Apache License 2.0
35 stars 2 forks source link

How to do an inference on my own dataset #4

Open JLinFreshForward opened 6 months ago

JLinFreshForward commented 6 months ago

Hi! Great work!

I am wondering how I can run a test on my own images?

lhaippp commented 6 months ago

Hi,

the whole process is: images -> stichting image -> rectangling image, for the first step, it is needed to stich images by yourself, e.g, using [UDIS2](https://github.com/nie-lang/UDIS2)

then we will recently upload a script to perform the second step, for given a already stichted image

wudami commented 6 months ago

Hello, how to generate mesh and mask for the stitched image?

luhan225 commented 2 months ago

Hi,

the whole process is: images -> stichting image -> rectangling image, for the first step, it is needed to stich images by yourself, e.g, using [UDIS2](https://github.com/nie-lang/UDIS2)

then we will recently upload a script to perform the second step, for given a already stichted image

Hello, may I ask when the script can be provided？

lhaippp commented 3 weeks ago

Hello, how to generate mesh and mask for the stitched image?

if you have an already stiching image without M_s, you could consider using this function define_mask_zero_borders to generate one mask

lhaippp commented 3 weeks ago

Hi, the whole process is: images -> stichting image -> rectangling image, for the first step, it is needed to stich images by yourself, e.g, using [UDIS2](https://github.com/nie-lang/UDIS2) then we will recently upload a script to perform the second step, for given a already stichted image

Hello, may I ask when the script can be provided？

sry for the delay, maybe we could find some time after cvpr2025 to finish the script, or we are happy to receive a PR