Catergorical classification with DNN and CNN.
Explore the repository»
View Problem Statement
View Report
tags : classification, categorization, fashionmnist, dnn, cnn, deep learning, tensorflow
This project is an implementation of the task of Fashion MNIST classification using the popular FashionMNIST dataset comprising of Zalando’s article images. We are required to identify an image and classify it to one of the ten available categories. The following models were implemented and the performance was evaluated.
- Multi-layer Neural Network
- Convolutional Neural Network
This project was built with
- python v3.7
- tensorflow v2.1
- The list of libraries used for developing this project is available at requirements.txt.
Clone the repository into a local machine using
git clone https://github.com/vineeths96/FashionMNIST
Please install required libraries by running the following command (preferably within a virtual environment).
pip install -r requirements.txt
We will load the Fashion MNIST using Tensorflow API. Hence no manual setup is necessary for the program.
The main.py
is the interface to the program. It is programmed to run in two modes – train mode and test mode. The main.py
file takes an optional command line argument to specify the mode of execution – whether to train or test model. The main.py
, when executed without any arguments, enters into testing the deep models, and produces the output files multi-layer-net.txt
and convolution-neural-net.txt
. The main.py
when executed with the (optional argument) --train-model
enters into training mode and saves the models after training.
python main.py --train-model
python main.py
Detailed discussions on results can be found in the report here
CNN | MLNN |
---|---|
![]() |
![]() |
![]() |
![]() |
Test Accuracy: 91.76% | Test Accuracy: 89.06% |
|
Distributed under the MIT License. See LICENSE
for more information.
Vineeth S - [email protected]
Project Link: https://github.com/vineeths96/FashionMNIST
-
Fashion-MNIST
Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms. Han Xiao, Kashif Rasul, Roland Vollgraf. arXiv:1708.07747