This is my independent coursework on deep learning and computer vision done as a postgraduate student at Queen Mary University.
-
The first course assignment is an experimental study of SRCNNs, a class of techniques used to enhance or increase the resolution of imaging systems. The convolutional neural network for image super-resolution is called SRCNN, which shows that traditional sparse coding-based SR methods can be reformulated as deep convolutional neural networks
-
The second task of the course is the study of GANs. Generative adversarial networks can be used to generate real data, such as images, and are divided into generative and discriminative models. The generative model is responsible for generating data that can confuse the discriminative model, while the discriminative model makes judgements based on real data.
-
The third task is to compare learning VGG and ResNet for image classification in terms of computational complexity, number of parameters, training time, and accuracy of test results respectively.