Sketch for multiple camera live stitching setup

[shuckerino]

Hello, as I have already shared in another discussion, I need to realise live video stitching with multiple (4) cameras. The plan is to mount 4 cameras on top of a model car within a 90 degree angle. Each camera has a field of view of around 110 degrees and they are positioned beside each other on top of the roof. I now want to develop a python script that can stitch the images of the 4 cameras together with a decent frame rate (the optimal goal would be 30 fps).
At the end I want to have a 360 degree panorama image (video) that I can live stream to UnrealEngine (there It will be displayed and used vor VR)

The following is a rough sketch (hope it is enough to display our idea):

My most important question is if this repository could technically handle this kind of a load (because its Python) or if this would not work with a decent frame rate (just want to hear some opinions or experiences from you). Otherwise I need to look for alternatives (C++ etc).

Thanks for your help!

[lukasalexanderweber]

Do you already have the cameras? Could you share test images of the setup? Then you could get some Timing Benchmarks. OpenCV is C++ and this is only a wrapper (running C in the Background), so it wont get that much faster.

Also check out #89, there are some problems integrating results in VR/360° Environments

[shuckerino]

Unfortunately, I have just been assigned to this task and dont have the real setup available. My current task is to just try to make live image stitching work for 2 webcams (but of course I want to make it already scalable to the 4 cameras). I will try to get test images from the real setup ASAP.

Good to know that C++ would not be much faster because of OpenCV basically being C(++).

And also thanks for mentioning the related issue, I will take a look at it and maybe come back with some more information later.

[lukasalexanderweber]

@shuckerino with ad961e8 it's now possible to stitch loaded images directly