3.5 Convolutional Neural Networks

1. Convolutional Neural Networks

In this video David explains the principles behind neural network, and introduces Convolutional Neural Networks, a type of deep learning network designed for image analysis.

Introduction: in the first presentation Dr. David Millard introduces and gives a brief explanation of the the remarkable advances which have been made in the field of image analysis due to developments in the use of feed-forward and feed-backward neural networks.


Neural Networks & Images: moving from pixel analysis to higher feature generation and analysis



Convolutions: from pixels to features - use of Convolutional Neural Networks



Filters: how do we determine which weightings and filter values will produce the 'best' results?



Summary: In this final presentation of the session Dave summarises the topics covered, explaining the pro's and con's of using Neural Networks.



The goal with this section is not to fully understand neural networks, but to take away some of the mystique, and to get across some of the main principles:

  • How neural networks become effective through training
  • How convolutions work to identify features in images using filters
  • How convolutions can be layered to identify higher level features from lower level ones
  • How convolutions can be included in the training process, learning filters automatically