trt-yolo-app aspect ratio for model other than 1:1

Is there a specific reason why you cannot use an aspect ratio for the yolo model in trt-yolo-app other than 1:1? For my purpose it would be advantageous to be able to set the ratio to 16:9 but assertions fire when you attempt to do that. Additionally, if anyone has had success being able to modify the code to work around these assertions could you please share?


This sample targets for demonstrating the usage of TensorRT and deepstream.
It won’t coverage all the possibility.

But since the implementation is totally open-sourced, you can still customize it for 16:9 aspect ratio.