Cropping images based on the predicted bounding boxes

Hey, guys!

I’m currently working on a object detection project and the code is running well!
The problem that I’m facing is with cropping the final predictions. I labeled the dataset in the COCO format but the regression results come from -1 to 1 and I just can’t figure out how to crop an image based on these results!!
I’ve searched quite a lot for the Flowfield conversion method that I read is used by the algortithm to find out how to convert it back to the COCO coordinates and crop with OpenCV.
Does anybody know how to proceed?

Have you found the solution?

Yes! This was quite a long time ago, but all I did was rescale the x and y coordinates resultant from the regression by the dimensions of the image!