Skip to main content
. 2023 Sep 20;63(19):6053–6067. doi: 10.1021/acs.jcim.3c00422

Figure 6.

Figure 6

Architecture of the arrow-extraction model. The model takes in an image patch which contains a single, isolated connected component. Features are extracted by the Resnet-18 backbone layers. Transformed features are fed to two parallel branches: one for arrow detection, and another for classification of arrows into arrow types (solid, curly, equilibrium, resonance).