Author: Vineet Jacob Kuruvilla, PhD
Date: 8th March 2020
The purpose of this project is to show an example of how to use Convolutional Neural Network for image classification tasks. In this project, I have used dog and cat image dataset to classify whether the image is that of a dog or a cat. There are two way to go about this task- 1) Transfer Learning ie., to use a pretrained network and, 2) Design your own CNN. I have used the approach #2 in this project.
Beginners
-
Using datastore
-
Augmenting image data for produce more training data than originally available
-
Defining a CNN architecture (here only the programmatic way is shown. I used the Deep Network Designer App to build the network. It saves a lot of time.)
-
Setting the training parameters (also enabling parallel computing option)
-
Finally, testing the accuracy of the trained network using test data
MATLAB 9.7
Deep Learning Toolbox 13.0
Machine Learning and Statistics 11.6
Parallel Computing Toolbox 7.1