Diabetes Noninvasive Recognition via Improved Capsule Network

Noninvasive recognition is an important trend in diabetes recognition. Unfortunately, the accuracy obtained from conventional noninvasive methods low. This paper proposes a novel Diabetes Recognition method via plantar pressure image and improved Capsule Network (DNR-CapsNet). The input of proposed image, output result: healthy or possibly diabetes. ResNet18 used as backbone convolutional layers to convert pixel intensities local features DNR-CapsNet. Then, PrimaryCaps layer, SecondaryCaps DiabetesCaps layer are developed achieve semantic fusion locality-constrained dynamic routing also further improve our method. experimental results indicate that has better performance on than state-of-the-art methods.