What is the efficient way to constract bounding boxes for an objeject given the encoded pixel of that object. I am trying to convert a task from segmentation to object detection
