Figure 4
From: Automatic choroidal segmentation in OCT images using supervised deep learning methods
Accuracy vs. Speed vs. Complexity comparison of the patch-based methods and a semantic segmentation method using the standard U-net architecture. The different network architectures and patch sizes are compared as is the use of contrast enhancement. Here, the accuracy (CSI mean absolute error in pixels [x-axis]) is compared to the network complexity (number of parameters [shape size]) and the evaluation time (per B-scan in seconds [y-axis]). Due to high similarity in accuracy, speed and parameters, only a single semantic segmentation architecture (standard U-net) is illustrated here.
