mortezamardani Goto Github PK

cs231n

Stanford CS231n Convolutional Neural Networks for Visual Recognition Assignments

dvh-prediction

Implemention of DVH prediction from the (contoured) anatomical scans ...

gan-hallucination

This projects investigates the possible hallucinations or adversarial attacks for solving linear inverse problems. The goal is to understand the possible hallucinations, define metrics to quantify the hallucination, and find regularization techniques to make deep reconstruction nets robust against hallucination.

graphautoencoder

Embedding the graph nodes using a deep fully connected autoencoder

mardani

modulus

A PyTorch based deep-learning toolkit for developing DL models for physical systems

neural-pgd

This code implements the neural proximal gradient descent (PGD) algorithm proposed in https://arxiv.org/abs/1806.03963. The idea is to unroll the proximal gradient descent algorithm and model the proximal using a neural network. Adopting residual network (ResNet) as the proximal, a recurrent neural net (RNN) is implemented to learn the proximal. The code is flexible to incorporate various training costs such as pixel-wise l1/l2 cost, adversarial GAN, LSGAN, and WGAN.

single-layer-cae

Implementation of CAE for dimensionality reduction of MNIST dataset

unpaired-gancs

This code implements the recovery of image x from the undersampled measurements y when the pair (x,y) is not avaialble for training. However, we know the images, say \tilde{x}, from another relevant image dataset that can guide us to recover the ground-truth x from y. This is important for medical imaging applications where usually one doesn't have access to high-resolution datastes for all organs.