Machine Visual Acuity | Brooke K. Ryan

[giscus[bot]]

Machine Visual Acuity | Brooke K. Ryan

In this project, we aim to gain insights about human visual acuity by applying these tests to machines. The main goal is to first train a convolutional-based neural network to recognize optotypes with low amounts of distortions so that it can use its knowledge to classify an unseen optotype from a testing set with optotypes with medium to high amounts of distortions. We used transfer learning, with the use of a VGG network, to obtain a baseline model for the problem. Then, we experimented with mixing the testing set with the training set, to determine if that could help the network make better predictions.

https://brookekryan.com/research/machine-visual-acuity.html