How many specimens make a sufficient training set for automated 3D feature extraction?
J. M. Mulqueeney, A. Searle-Barnes, A. Brombacher, M. Sweeney, A. Goswami, T. Ezard
This study determines the minimum number of training images needed for accurate 3D feature extraction using deep learning and evaluates the effectiveness of data augmentation.
bioRxiv · Jan 10, 2024 · PreprintEvolutionary Biology