Energy Efficient Learning Algorithms for Glaucoma Diagnosis

Code

• github.com/KrishNachnani/glaucoma-detector-ai
  tf★ 2

Abstract

Glaucoma is a widespread issue that affects millions of individuals across the globe. A shortage of trained ophthalmologists is just one reason why broad screening is challenging, particularly in rural areas. To address this issue requires not just automation assistance but efficient assistance to match the limited resources in rural areas. This study evaluates energy efficient techniques to classify glaucoma -- techniques that can enable efficient automation assistance for physicians operating in limited environments. Two energy effective detection techniques (MobileNetV2 and machine learning algorithms over featurized data) are evaluated on the ORIGA dataset and the results are compared to variants of ResNet, a state-of-the-art convolutional neural network. While the featurization method exceeds ResNet's performance, MobileNetV2 falls short. This work provides a pathway for energy efficient implementations of glaucoma detection.

Reproductions