A Deep Learning course project at the University of Oulu: detecting diabetic retinopathy from retinal images, computer vision applied to healthcare diagnostics.
I developed a deep learning model for the task, experimenting with different neural network architectures and optimization techniques. The repository walks through the comparisons, including Grad-CAM visualizations, attention mechanisms and ensembles.
Rather than training from scratch, I used transfer learning on pretrained convolutional networks including ResNet and VGG, adjusting the model weights for the task, and compared them against each other.