0
$\begingroup$

Is it possible to get a fixed number of boxes/instances from Faster/Mask R-CNN algorithms?

I am using Mask R-CNN to segment 5 layers in a retina OCT image. But it only produces 2-3 layers, even though all the input images in the training set have 5 layers annotated.

Is there a way to force a fixed number of instances?

$\endgroup$

1 Answer 1

0
$\begingroup$

I am using Detectron2 library to run Faster/Mask R-CNN. One can set a maximum for the number of detected objects by setting the parameter NMS_THRESH_TEST. However, there is no way to put a lower bound for the number of detected objects. One can only fine tune this parameter SCORE_THRESH_TEST to get more objects detected though.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.