參賽者學校

嘉諾撒聖瑪利書院

理念解構

OptiScan is an AI system which analyses retinal images to detect potential glaucoma patients. Our system is trained in Colab. Using the transfer learning method, we input a dataset of 2333 retinal images into a pre-trained ResNet-50 model. The dataset consists of retinal images of glaucoma patients and normal people. We used Tensorflow, together with PyTorch, as the library for the dataset. 80% of the dataset is used to train our system, while the remaining 20% acts as the validation data. In the end, our system has an accuracy of 88.2%.