Fashion Mnist Classifier
BHUMIK KUMAR KAPOOR
New Delhi, Delhi
- 0 Collaborators
This project implements a multi-class image classifier for the Fashion MNIST dataset using a Convolutional Neural Network (CNN) and Streamlit for the web interface. Users can upload images to classify them into one of ten fashion categories. ...learn more
Project status: Under Development
Intel Technologies
Other
Overview / Usage
This project implements a multi-class image classifier for the Fashion MNIST dataset using a Convolutional Neural Network (CNN) and Streamlit for the web interface. Users can upload images to classify them into one of ten fashion categories.
Methodology / Approach
Model ArchitectureThe CNN architecture consists of the following layers:
- Convolutional Layer: Applies 16 filters of size 3x3 with ReLU activation and padding.
- Max Pooling Layer: Reduces dimensionality with a kernel size of 2x2.
- Fully Connected Layer: Outputs class predictions for the 10 categories in Fashion MNIST.
Uploaded images are preprocessed using the following steps:
- Convert the image to grayscale.
- Resize the image to 28x28 pixels.
- Normalize the pixel values to be between -1 and 1.
Technologies Used
o run this project, you'll need the following Python libraries:
streamlit
torch
torchvision
torchmetrics
Pillow
numpy
You can install the required packages using pip:
pip install streamlit torch torchvision torchmetrics Pillow numpy