Skip to main content
. 2023 Sep 26;39(10):btad594. doi: 10.1093/bioinformatics/btad594

Figure 2.

Figure 2.

Schematic diagram of the model architecture of ScribbleDom, which consists of two 1 × 1 convolutions and two inception blocks. The inception blocks facilitate capturing features at different resolutions. Loss is calculated using cross entropy between the predicted cluster label and the scribble (for the scribbled spots) or the response map (for the spots without scribbles).