Aim: The goal is to build a model which classifies the type of an unseen image as accurate as possible, by implementing, evaluating, and comparing amongst 2 different multi-layer perceptron Neural Networks.
Datasets (from https://web.engr.oregonstate.edu/~tgd/bugid/stonefly9/)
- YOR Dataset - containing 483 images of Yoraperla type.
- CAL Dataset - containing 459 images of Calineuria type.
- Please check my notebook with the name
CSI5341 Assignment 2 - Kelvin Mock 300453668(in either.pdfor.ipynbformat) to see the main flow of my analysis. - Multi-Layer Perceptron (MLP) model - please check
MLP.py. - Convolutional Neural Network (CNN) model - please check
CNN.py. - Use of Keras API of Tensorflow
- Involves some logic to laod images and to perform image preprocessing steps on them - please check
setup.py. - Constants file which contains all kinds of configuration, such as the directories to the original datasets, size of the test set, and specific parameters for each neural network.
- Models are exported into
.pkl filesduring each stage of my modelling and analysis work.