Module Catalogues, Xi'an Jiaotong-Liverpool University   
Module Code: EEE408
Module Title: Deep Learning in Computer Vision
Module Level: Level 4
Module Credits: 5.00
Academic Year: 2017/18
Semester: SEM2
Originating Department: Electrical and Electronic Engineering
Pre-requisites: N/A
Computer Vision, a core topic in multimedia, has become ubiquitous in our society, with applications in search, image understanding, apps, mapping, medicine, drones, self-driving cars, and even communication systems. Core to many of these applications are visual recognition tasks such as image classification, localization and detection. Recent developments in neural network approaches (“deep learning”) have greatly advanced the performance of these state-of-the-art visual recognition systems. This course focusses on the details of these deep learning architectures diving deeply into the end-to-end models for these tasks, particularly image classification. During this course, students will learn to implement, train and debug their own neural networks and gain a detailed understanding of cutting-edge research in computer vision and multimedia applications.
Learning outcomes 
Demonstrate expert knowledge of and/or be able to offer critical insight into the:

A field of machine learning and computer vision, from history and development to background theory and system architecture,

B current state of the art in Deep learning and Deep understanding in relation to machine learning,

Show the intellectual ability to provide:

C critical analysis how deep learning and deep understanding relate to other existing machine learning methodologies and techniques, for example, computer vision pipelines,

D an analysis of real world data problems and design suitable solutions based on available technologies,

E an analysis of the performance of deep learning algorithms,

Demonstrate the following module specific practical skills:

F creation of neural networks,

Demonstrate the following general transferable skills:

G undertake individual research on current state of the art technologies.

Method of teaching and learning 
This module will be delivered through a combination of formal lectures and lab sessions. The lab experiments will be performed using C/C++, MATLAB or PYTHON or other software tools for deep learning. For each lab a report shall be prepared, two of these will count towards the summative assessment for the module. A final examination consisting of a number of problem/design based questions as well as questions aimed at examining the students grasp of the subject theory will form the remainder of the summative assessment.
1. Introduction to computer vision and deep learning, historical context.

2. Image classification and the data-driven approach

k-nearest neighbor

Linear classification

Higher-level representations, image features

Optimization, stochastic gradient descent

3. Back-propagation

Introduction to neural networks

4. Training Neural Networks Part 1

activation functions, weight initialization, gradient flow, batch normalization

babysitting the learning process, hyperparameter optimization

5. Training Neural Networks Part 2: parameter updates, ensembles, dropout

Convolutional Neural Networks: introduction

6. Convolutional Neural Networks: architectures, convolution / pooling layers

Case study of ImageNet challenge winning ConvNets

7. ConvNets for spatial localization

Object detection

8. ConvNets for image segmentation

9. Understanding and visualizing Convolutional Neural Networks

10. Recurrent Neural Networks (RNN), Long Short Term Memory (LSTM)

RNN language models

Image captioning

11. Advanced topic of other deep learning models e.g., Deep Reinforcement Learning, Restricted Boltzmann Machines (RBM) based DNN.

12. Overview of Caffe

13. Summary
Delivery Hours  
Lectures Seminars Tutorials Lab/Prcaticals Fieldwork / Placement Other(Private study) Total
Hours/Semester 26      24    100  150 


Sequence Method % of Final Mark
1 Lab Reports 20.00
2 Lab Reports 20.00
3 Final Exam 60.00

Module Catalogue generated from SITS CUT-OFF: 10/22/2017 10:48:12 AM