The "laneBoundaryDetector" that loads the pretrained "CLRNet" by default already applies ROI using the "detect" function as follows, which is also seen in the example you referenced above:
detector = laneBoundaryDetector;
lanes = detect(detector,I,ROI=120,ExecutionEnvironment="cpu");
The "ROI" parameter either crops the top 120 rows of the image in this snippet, or the function crops the image using the position and size of the crop rectangle specified as "[x y width height]". For more information, refer to this documentation for the "detect" function:
Hope this helps!
