Object detection with stitching

[meliioko]

Hello, I am working on a model of object detection, on mutli cameras so i would like to stitch the images of the cameras. The model gives back the coordinate of the box surrounding the object, and i would like to map those cordinates to the panorama given back by the Stitcher but I'm having trouble doing it. Has anyone worked on a similar task ?

[lukasalexanderweber]

You have the cooridinates in a single Image and plot them on the final stitched result? The key might be perspectiveTransform

See https://stackoverflow.com/questions/76003489/how-to-get-image-coordinates-after-successful-match-and-stitch-using-opencv

[lukasalexanderweber]

You can customize the workflow. See https://github.com/OpenStitching/stitching/blob/main/stitching/stitcher.py and https://github.com/OpenStitching/stitching_tutorial/blob/master/docs/Stitching%20Tutorial.md

[meliioko]

I have succeeded in retrieving the coordinates in the panorama. Thanks a lot. I wonder if it could be a feature we add to this project, maybe someone else will need it in the future, I guess multi camera tracking is not something niche.

[lukasalexanderweber]

I would be glad if you share your solution. We can put the samples here and can create a faq content with your code in the discussions or put it in a tutorial. See e.g. the VideoStitcher. While other people might want to stitch videos, I still think it's better when user create their own specific stitcher classes for their specific use case.

[lukasalexanderweber]

You can also demonstrate it with existing samples. If I understood correctly you had a Rectangle on a single input image:

and transformed the coordinates to plot the Rectangle on the final result

[lukasalexanderweber]

@meliioko where is your sollution? Don't you want to contribute?