Ecosyste.ms: Awesome
An open API service indexing awesome lists of open source software.
https://github.com/ranpy13/digit_recognition_rmodel
Implementing handwriting recognition for digits in a grayscale image grid stored in csv files, implemented using single layer and multi layer convolution neural networks through mxnet on R
https://github.com/ranpy13/digit_recognition_rmodel
Last synced: 6 days ago
JSON representation
Implementing handwriting recognition for digits in a grayscale image grid stored in csv files, implemented using single layer and multi layer convolution neural networks through mxnet on R
- Host: GitHub
- URL: https://github.com/ranpy13/digit_recognition_rmodel
- Owner: ranpy13
- License: mit
- Created: 2024-11-09T07:48:09.000Z (8 days ago)
- Default Branch: main
- Last Pushed: 2024-11-09T11:38:01.000Z (8 days ago)
- Last Synced: 2024-11-09T12:25:06.327Z (8 days ago)
- Language: R
- Size: 18.9 MB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Code of conduct: CODE_OF_CONDUCT.md
Awesome Lists containing this project
README
# Handwritten Digit Recognition using Convolutional Neural Networks
This project implements a handwritten digit recognition system using various machine learning techniques, including logistic regression, single-layer neural networks, multi-layer neural networks, and convolutional neural networks (CNNs). The dataset used is the popular MNIST dataset, which consists of 28x28 pixel images of handwritten digits (0-9).
## Table of Contents
- [Introduction](#introduction)
- [Data Exploration](#data-exploration)
- [Modeling Techniques](#modeling-techniques)
- [Logistic Regression](#logistic-regression)
- [Single-layer Neural Networks](#single-layer-neural-networks)
- [Multi-layer Neural Networks](#multi-layer-neural-networks)
- [Convolutional Neural Networks](#convolutional-neural-networks)
- [Results](#results)
- [Visualizations](#visualizations)
- [Installation](#installation)
- [Usage](#usage)
- [License](#license)
## Introduction
The goal of this project is to classify handwritten digits using deep learning techniques. The project explores different models and evaluates their performance on the MNIST dataset.
## Data Exploration
The dataset is loaded from a CSV file, and the following steps are performed:
- Dimensions and structure of the data are examined.
- Unique labels and pixel value ranges are analyzed.
- Sample images are visualized.
- The target variable is transformed from integer to factor for classification.
```r
data <- read.csv("train.csv")
dim(data)
head(data[1:6])
summary(data$label)
```
## Modeling Techniques
### Logistic Regression
A multinomial logistic regression model is trained and evaluated.
```r
library(nnet)
model_lr <- multinom(label ~ ., data=data_train)
prediction_lr <- predict(model_lr, data_test, type = "class")
```
### Single-layer Neural Networks
A single-layer neural network is implemented using the `nnet` package.
```r
model_nn <- nnet(label ~ ., data=data_train, size=50)
prediction_nn <- predict(model_nn, data_test, type = "class")
```
### Multi-layer Neural Networks
A multi-layer neural network is built using the `mxnet` package.
```r
require(mxnet)
model_dnn <- mx.model.FeedForward.create(softmax, X=data_train.x, y=data_train.y)
```
### Convolutional Neural Networks
A convolutional neural network is constructed to improve classification accuracy.
```r
conv1 <- mx.symbol.Convolution(data=data, kernel=c(5,5), num_filter=20)
model_cnn <- mx.model.FeedForward.create(softmax, X=train.array, y=data_train.y)
```
## Results
The performance of each model is evaluated using confusion matrices and accuracy metrics.
```r
cm_lr <- table(data_test$label, prediction_lr)
accuracy_lr <- mean(prediction_lr == data_test$label)
```
## Visualizations
Visualizations include:
- Sample images from the dataset.
- Activation maps from convolutional layers.
- Learning curves for model training.
```r
par(mfrow=c(4,4))
for (i in 1:16) {
outputData <- as.array(executor$ref.outputs$activation15_output)[,,i,1]
image(outputData, xaxt='n', yaxt='n', col=grey.colors(255))
}
```
## Installation
To run this project, you need to have R and the following packages installed:
- `nnet`
- `caret`
- `mxnet`
You can install the required packages using:
```r
install.packages(c("nnet", "caret"))
```
For `mxnet`, follow the installation instructions on the [MXNet R package page](https://mxnet.apache.org/get_started/).
## Usage
1. Clone the repository:
```bash
git clone https://github.com/ranpy13/digit_recognition_Rmodel.git
cd digit_recognition_Rmodel
```
2. Load the dataset and run the R script:
```r
source("main.R")
```
## License
This project is licensed under the MIT License. See the [LICENSE](LICENSE) file for details.
---
Feel free to modify the content as needed to better fit your project specifics!