Is it possible to perform instance segmentation if you have the following:
- Binary Segmentation Map
- Bounding Boxes (with respective class)
Let's say we're doing something within cellular microscopy and we have a segmentation map showing us the cell masks. We also have run an object detection model and generated boxes and classes for each cell in the image. Assume some of the cells are overlapping (segmentation map is not sufficient by itself to do instance segmentation... even binary). Can we take this information and use it to generate an instance segmentation map?
Is this a possible task?