Applications of Deep Learning
-
The first project consisted in a convolutional implementation where we learned how to work with padding and no-padding and how to apply some filters like Sobel.
-
The second project consisted in an implementation of the VGG16 architecture using a dataset that have chest-x-rays. The aim was to replicate the analysis made in the article, but using vgg16 and comparing the results using pooling operation ( None, avg max ), weights initialization and transfer learning.
-
The third part of the project consisted in the improvement of the last model ( vgg16 ), using batch normalization, convolutional reduction, skip connections, factored convolution and finally testing with DenseNet architecture.
-
Dataset explanation: https://aimi.stanford.edu/chexpert-chest-x-rays
-
Article: https://arxiv.org/abs/1901.07031
-
Dataset from kaggle: https://www.kaggle.com/datasets/mimsadiislam/chexpert