Review: Mask R-CNN Models

Document Type : Original Article


Faculty of Artificial Intelligence, Kafrelsheikh University, Kafrelsheikh, Egypt


Instance segmentation is a challenging computer vision task that requires the prediction of object 
instances and their per-pixel segmentation mask. This makes it a hybrid of semantic segmentation and 
object detection. It detects and delineates each distinct object of interest appearing in an image. Mask RCNN model is common for instance segmentation that has several versions for improving this task. We 
proposed a simple comparison between Fifteenth different version frameworks from Mask-RCNN for 
object instance segmentation. Our survey representing the difference between the popular versions of 
Mask R-CNN. The Mask R-CNN method extends Faster R-CNN by adding a branch for predicting an 
object mask in parallel with the existing branch for bounding box recognition. The results in most 
versions were implemented on of the COCO dataset that created for instance segmentation tasks.