From object segmentation to instance segmentation

First, we will present some tools that we can use to obtain instance masks from the segmentation models we just covered. The U-Net authors popularized the idea of tuning encoders-decoders so that their output can be used for instance segmentation. This idea was pushed further by Alexander Buslaev, Victor Durnov, and Selim Seferbekov, who famously won Kaggle's 2018 Data Science Bowl (https://www.kaggle.com/c/data-science-bowl-2018), a sponsored competition to advance instance segmentation for medical applications.