Using MaskRCNN for only detection (not instance segmentation)

Is there a way to use MaskRCNN for only detection labels and for detection bounding-box prediction? All the tutorials I’ve seen so far seem to require segmentation mask labels as well. My problem only has bounding box labels and only needs bounding box predictions.

Hi Lakshay!

Not that I’m aware of (without modifying it significantly).

Consider using Faster R-CNN which is designed for detection rather
than for instance segmentation. (And it is pretty close to being the
bounding-box piece of Mask R-CNN.)


K. Frank