Mask RCNN is Faster RCNN (object detection with bounding boxes) with a mask on it. It is developed by Facebook AI Research (FAIR).  The output from Mask RCNN is similar to YOLO (You Only Look Once) but the implementation strategy is different.

What does mask do in Mask RCNN?

Mask features labels each pixel and compares each pixel with an object.

Here is the link of the paper written the Facebook AI Research (FAIR) on Mask RCNN. Feel free to go through the above link for more detail. If you want to see quick video from the creator which was presented in ICCV17 you can find the like here. In this video, he has mentioned how far we have reached in object detection and semantic segmentation but we have not achieved much in Instance segmentation.

Applications of Mask RCNN

Joint Object Detection
Pose Estimation
Instance Segmentation