Evaluation of CNN Models with Fashion MNIST Data

Thumbnail Image
Date
2019-01-01
Authors
Zhang, Yue
Major Professor
Joseph Zambreno
Advisor
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Authors
Research Projects
Organizational Units
Journal Issue
Is Version Of
Versions
Series
Department
Electrical and Computer Engineering
Abstract

The work seeks to evaluate the performance of four CNNs with respect to Fashion MNIST data set. Fashion MNIST is a dataset of images consisting of 70000 28*28 grayscale images, associated with a label from 10 classes. In this report, the accuracy of four popular CNN models that are LeNet-5, AlexNet, VGG-16 and ResNet for classifying MNIST-fashion data revealed that ResNet was the best suited for the selected dataset. The training process has been coded with Tensorflow. After the result accuracy improving, we could use the new model to the fashion company that can help the fashion company more accurately classify clothing. Moreover you could build your own closet online for your fashion.

Comments
Description
Keywords
Citation
DOI
Source
Subject Categories
Copyright
Tue Jan 01 00:00:00 UTC 2019