Machine Learning: Vision

Jan 10th, 2022

Generative adversarial network unlike other classes of machine learning frameworks trains a network in an unsupervised manner using a generator and a discriminator. The generative network generates new data from a training set with similar statistics while the discriminative network evaluates them and dynamically updates itself. This article explores state-of-the-art GANs, existing applications, promising future directions and inherent concepts of GAN algorithm.

Transformers in Vision


Computer vision community has recently shown a strong interest in the transformer network, which was initially developed by the Natural Language Processing (NLP) community. This article will discuss the prominent concepts in transformer networks, state-of-the-art transformers, applications of transformers in vision, the advantages and limitations of such transformers.

Learning Imaging


Imaging has always been human-centred, capturing images by understanding the environment and tuning different parameters with experience to capture candid shots. While it is similar to photography, imaging in robots have more control over the environment and imaging system over a human. This article will discuss how a robot can learn to perform imaging and future avenues of research in this domain.

%d bloggers like this: