Module Code - Title:
CE5021
-
DEEP LEARNING FOR COMPUTER VISION
Year Last Offered:
2023/4
Hours Per Week:
Grading Type:
N
Prerequisite Modules:
Rationale and Purpose of the Module:
Deep learning has become the dominant approach to designing solutions for many common computer vision tasks. In this module the application of deep learning to the key computer vision tasks of image classification, object detection, semantic segmentation and facial recognition is discussed in detail. Fundamental concepts in the design and structure of deep neural networks will also be discussed, so students gain a full understanding of how to design and build networks for their own applications.
Syllabus:
Introduction to Deep Learning
Convolution Neural Networks (Padding, Pooling, Receptive Field, Convolution with multiple kernels)
Image Classification with Deep CNNs (AlexNet, VGG, GoogLeNet)
Advanced Networks for Image Classification (ResNets, SE-Net, DenseNet)
Training Deep Networks with Keras/Tensorflow/Pytorch
Visualising Neural Networks
Transfer Learning & Applications
Region Proposals
Networks for Object detection (RCNN & Derivatives)
Single Stage Object detection (Yolo, SSD)
Semantic Segmentation (Full Convolutional Networks, Transpose Convolution, DeepLab)
Introduction to Facial Recognition
Metric learning for facial recognition with DNNs.
Learning Outcomes:
Cognitive (Knowledge, Understanding, Application, Analysis, Evaluation, Synthesis)
On successful completion of this module, students will be able to:
Demonstrate an understanding of a deep learning approach compared to classical approaches to image classification and object detection.
Demonstrate an understanding of recent advances in neural networks for image classification, object detection, semantic segmentation, and other tasks
Demonstrate an understanding of the application of transfer learning to developing deep learning based systems for object detection, semantic segmentation and facial recognition.
Demonstrate an ability to design and implement neural networks for computer vision tasks using the Tensorflow Keras or Pytorch APIs.
Affective (Attitudes and Values)
On successful completion of this module, students will be able to:
Given a computer vision problem, identify and defend an appropriate technique to tackle it.
Contribute meaningfully to an engineering team project development on deep learning based computer vision systems.
Psychomotor (Physical Skills)
On successful completion of this module, students will be able to: n/a
How the Module will be Taught and what will be the Learning Experiences of the Students:
The module will be delivered in an online and/or blended learning format, with online pre-recorded video lectures and interactive Q+A tutorials. The students will also complete a series of coding exercises either in a supervised lab setting (Full time cohort) or in their own time with online moderator support (Part Time cohort).
Research Findings Incorporated in to the Syllabus (If Relevant):
Prime Texts:
Russell, Stuart J. and P. Norvig (2016)
Artificial Intelligence: a modern approach
, Pearson Education Limited, 2016
Other Relevant Texts:
Programme(s) in which this Module is Offered:
PDCVSYTPA - Computer Vision Systems
Semester - Year to be First Offered:
Module Leader:
Tony.Scanlan@ul.ie