I recently go interested in image segmentation. To understand the work going on in this field, I decided to look at some of the papers tackling semantic segmentation and instance-aware semantic segmentation. I compiled what I read and learned in this blog:
I was astounded by some of the research being done! Although, methods for semantic segmentation simply use an encoder-decoder framework based on FCNs with changes usually only being made in the decoder network, many different approaches have been taken to solve the instance segmentation task. I especially found the work on InstanceFCN and FCIS, where they find instance sensitive score maps, really interesting and ingenious.
What other interesting works on segmentation have you seen and come across? Also, which approach to segmentation is better? FCN based, like that in FCIS or detection based, like in Mask R-CNN?