Hi everyone a couple weeks ago I asked some questions about image stitching for a soccer field. At the start I was using python but we switched over to cpp. After pouring in a shit ton of hours we came to a pretty good/okay sollution. The images are stitched, the video is exported and we can play it on for example a website. The program is also pretty fast, the input (video’s/frames) have an fps of 25 and both 4000x3000 relosution. But we can process them in around 0.024 seconds with an export to 1080p.
Here is an foto of the field stitched, and unstitched.
Its pretty okay for an image manipulated by hand I think.
But everyone: Here is the github repo (its WIP so the code is still a bit messy):
If you see any improvements or mistakes please leave a comment, or use the code yourself!
Tested with:
Windows 10
Opencv 4.8-dev
Cuda arch 7.5 RTX 2060
Cuda toolkit 12.2
CUDNN (ver 8.9.5)
Nvidia Codec https://developer.nvidia.com/nvidia-video-codec-sdk/download~
I know there is a lot of room for improvement, but I’m pretty pleassed with what we managed to make in just a few weeks of evening hours.