Fine-tuning Convolutional Neural Networks: a comprehensive guide and benchmark analysis for Glaucoma Screening

Abstract

This work aimed at giving a comprehensive, indetailed and benchmark guide on the route to fine-tuning Convolutional Neural Networks (CNNs) for glaucoma screening. Transfer learning consists in a promising alternative to train CNNs from scratch, to avoid the huge data and resources requirements. After a thorough study of five state-of-the-art CNNs architectures, a complete and well-explained strategy for fine-tuning these networks is proposed, using hyperparameter grid-searching and two-phase training approach. Excellent performance is reached on model evaluation, with a 0.9772 AUROC validation rate, giving arise to reliable glaucoma diagnosis-help systems. Also, a baseline benchmark analysis is conducted, studying the models according to performance indices such as model complexity and size, AUROC density and inference time. This in-depth analysis allows a rigorous comparison between model characteristics, and is useful for giving practioners important trademarks for prospective applications and deployments.

Wavelet Image Scattering Based Glaucoma Detection

“…This stems from the fact that they usually have lots of tunable hyperparameters that can only be tuned properly with large amount of training data which is not always readily available in biomedical data analysis problems [12]. Secondly, their optimal network architecture and hyperparameters configuration are not well understood [26]. Therefore, a simpler alternative method that will produce the same or even better result than CNNs will be desirable.…”