Figure 6
The diagram compares a Pretrained Model in the upper row with a Custom Model in the lower row. In both rows, an Input passes through four sequential feature-processing layers connected by right-pointing arrows. A horizontal bracket labelled Common Inner Layer spans these shared layers, indicating that they are reused in both models. In the pretrained model, the final feature layer connects to a fully connected network containing several node layers and ellipses for additional nodes, ending in an Output with multiple output nodes. In the custom model, the same common inner layers feed a newly configured fully connected network. A bracket labelled Custom Final Layers spans the custom network and its output section. The custom model ends with its own Output containing multiple output nodes.Architecture of a transfer learning