NNimageclassification
Neural Network for Image Classification final project for UCLA M156: Machine Learning, Spring 2017
• Code the neural network
• Define a training set and a test set
• Test several different designs of your network
• Evaluate performance in various scenarios (different classes to discriminate)
• Bonus: Pre-process the data with PCA to see how it affects performance and running time
Caltech 101, 28x28 Silhouettes
DUNLAP, LAUREN • HU, CARSON • LIU, JIACHEN
Instruction to run Convolutional Neural network files:
$cd CNN $python testCNN.py
Best Parameter: Conolution_1d 28 -> Maxpool -> Convolution_1d 56 -> Convolution_1d 224 -> Max_pool -> Fully_connected 1568 -> Fully_connected 101
Best result: 90.54%
required package: tensorlfow, tflearn, numpy, scipy