Distance to Center of Mass Encoding for Instance Segmentation
Thomio Watanabe, Denis F. Wolf · 2018
Instance segmentation can be considered an extension of the object detection problem where bounding boxes are replaced by object contours. Strictly speaking the problem requires to identify each pixel instance and class independently of the employed method. The advantage of instance segmentation over the usual object detection lies in the precise delineation of objects improving object localization. Additionally, object contours allow the evaluation of partial occlusion with basic image processing algorithms. This work approaches the instance segmentation problem as an annotation problem and presents a technique to encode annotations. We propose a mathematical representation of instances that any deep semantic segmentation model can learn and generalize. Each individual instance is represented by a center of mass and a field of vectors pointing to it. This encoding technique has been denominated Distance to Center of Mass Encoding (DCME).