CS 433 : Deep Learning & Computer Vision

Home CS 433 CS 439 VIRG

CS 433 : Deep Learning & Computer Vision

Autumn 2019

Recent developments in Deep Learning has not only helped computer scientists better understand about Images and Natural Language, generative neural network models are helping drug designers design potent drugs more efficiently. While deep learning has helped machines better interpret images than ever before, it is constantly creating new genres for research, development, training and applications. This course is designed to study how the field of computer vision – in the last decade – has evolved around ideas developed from years of research in deep learning. In this course, we made an effort to teach how to implement the cardinal aspects of what a deep learning & computer vision practitioner typically does in 2019.

Autumn 2019. We are recording the lectures as we are going along, once the course is complete, the lecture videos and slides will be uploaded for free public access.

Course Instructor : Sourajit Saha

Module 1 : Introduction to Convolutional Neural Networks

  • Lecture 1 : Introduction to Neural Networks & Deep Learning.
  • Lecture 2 : Loss Functions, Back-Propagation, Optimization.
  • Lecture 3 : Convolutional Neural Networks – how do they work.
  • Lecture 4 : Training a CNN in Pytorch to classify images.

Module 2 : Case Study – CNN architectures for Image Classification

  • Lecture 5 : VGG, Inception V3, Dense Net, Res Net, Res Net v2.
  • Lecture 6 : ResNext, Data Augmentation & Transfer Learning using Pytorch.
  • Lecture 7 : Neural Ordinary Differential Equations & why should we care about them.

Module 3 : Object Detection

  • Lecture 8 : Introduction to Object Detection, how to deal with Object Detection Datasets.
  • Lecture 9 : RCNN, Fast RCNN, Faster RCNN – How do they work.
  • Lecture 10 : RCNN, Fast RCNN, Faster RCNN – How to implement them.
  • Lecture 11 : YOLO, YOLOV3, SSD – How do they work.
  • Lecture 12 : YOLO, YOLOV3, SSD – How to implement them.

Module 4 : Dense Prediction

  • Lecture 13 : Introduction to Semantic Segmentation, how to deal with Semantic Segmentation Datasets.
  • Lecture 14 : FCN, U-Net, E-Net – How do they work.
  • Lecture 15 : FCN, U-Net, E-Net – How to implement them.
  • Lecture 16 : DeepLab V3+, ESP Net – How do they work.
  • Lecture 17 : DeepLab V3+, ESP Net – How to implement them.

Module 5 : Advanced Computer Vision Topics expedited by Deep Learning

  • Lecture 18 : Introduction to Generative Modeling.
  • Lecture 19 : Introduction to Generative Adversarial Networks(GANs), Implementing DCGAN in Pytorch.
  • Lecture 20 : Adversarial Attacks, DeepDream, Neural Style Transfer : In Pytorch.
  • Lecture 21 : Introduction to Recurrent Neural Networks.
  • Lecture 22 : Implementing LSTM, GRU : In Pytorch.
  • Lecture 23 : Soft Attention in Image Classification.
  • Lecture 24 : How to write a custom layer & a custom loss function in CNN : In Pytorch.