An Image Classification Algorithm built with data from the CIFAR-10 dataset.
Branch: master
Clone or download
Latest commit 9504497 Jan 14, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
README.md Update README.md Jan 14, 2018
batches.meta.mat Add files via upload Jan 14, 2018
classify.m Add files via upload Jan 14, 2018
displayData.m
featureNormalize.m Add files via upload Jan 14, 2018
fmincg.m
learningCurve.m Add files via upload Jan 14, 2018
loadData.m Add files via upload Jan 14, 2018
loadDataClassify.m Add files via upload Jan 14, 2018
main.m Add files via upload Jan 14, 2018
nnCostFunction.m Add files via upload Jan 14, 2018
param.mat Add files via upload Jan 14, 2018
param1.mat Add files via upload Jan 14, 2018
param2.mat Add files via upload Jan 14, 2018
param3.mat Add files via upload Jan 14, 2018
pickle.m Add files via upload Jan 14, 2018
predict.m Add files via upload Jan 14, 2018
predictTest.m Add files via upload Jan 14, 2018
randomInitTheta.m
sigmoid.m Add files via upload Jan 14, 2018
sigmoidGradient.m Add files via upload Jan 14, 2018
testY.m Add files via upload Jan 14, 2018

README.md

Image-Classification

An Image Classification Algorithm built with data from the CIFAR-10 dataset in Octave.

This program runs a feedforward neural network that received a 55% accuracy rate on a test dataset seperate from the training dataset. The neural network has 3 layers, the first having 1024 nodes as all input images are 32x32, the second having 81 nodes, and the last having 10 nodes corresponding to the 10 different classes.

Prior to processing images through the neural network, all images are reshaped into a grid, transformed into greyscale, and finally reshaped into rows, where each row represents an image in greyscale.

Training the neural network occurs in the main.m file.

Note that this code will not run properly. The dataset files were too large to upload to github but all of the code is here.