Description
This demo uses the initial drusen model that can be found here. The model is a simple convolutional neural network that takes in a 256x256 image and outputs a 256x256 image with the same number of channels. The model was trained on 262 images of drusen and 150 images of non-drusen. The model was trained for 100 epochs and the loss function was binary cross entropy. You can see all the objects types it has been trained to recognize in drusen_classes.js.
Testing Data
Need some test data? Try these: here.
Status
Model Output