Inference for detection on huge images

vision
1

Hi,
I trained a Mask R-CNN model on patches/crops 512x512 to detect an object.
Now I need to apply this model on huge images, 50K x 50K.
Is there a function that would take care of it? Because objects partially detected on the border will create problems.
Thanks in advance.

1 Like
2

It is called sliding / sliced / tiled inference.
The package SAHI: Slicing Aided Hyper Inference should do what you’re after.