Scribd
Upload
18 views13 pages

Handwritten Digit Recognition With CNN

The document presents a project on recognizing handwritten digits using Convolutional Neural Networks (CNNs), trained on the MNIST dataset. It outlines the CNN architecture, methodology, and results, achieving a model accuracy of 97.7%. The project emphasizes the importance of CNNs in image recognition tasks and their effectiveness in classifying handwritten digits.

Uploaded by

dwima1583
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
18 views13 pages

Handwritten Digit Recognition With CNN

The document presents a project on recognizing handwritten digits using Convolutional Neural Networks (CNNs), trained on the MNIST dataset. It outlines the CNN architecture, methodology, and results, achieving a model accuracy of 97.7%. The project emphasizes the importance of CNNs in image recognition tasks and their effectiveness in classifying handwritten digits.

Uploaded by

dwima1583
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 13

Handwritten Digit

Recognition using CNN

Presented By:
Dwima Guthal Basumatary
(202102021034)
Pamee Brahma (202102023127)
TABLE OF CONTENTS
INTRODUCTION MODEL ARCHITECTURE
01 ➔ Project overview
and objectives
04 ➔ Structure of CNN
Model

UNDERSTANDING CNN RESULTS


02 ➔ What is CNN and
why Convolution
05 ➔ Model accuracy
and sample
prediction
METHODOLOGY
03 ➔ Dataset and CNN
Working
06 CONCLUSION
INTRODUCTION
➔ This project focuses on developing a system for recognizing
handwritten digits using Convolutional Neural Networks
(CNNs).
➔ The model is trained on the MNIST (Modified National
Institute of Standards and Technology) dataset, a widely
used benchmark for image recognition tasks in deep learning.
OBJECTIVES
➔ Develop a CNN model capable of
accurately classifying handwritten digits
from 0 to 9.
➔ Show how handwritten digit recognition
helps in reading bank checks, sorting
mail, mobile number entry, and license
plate recognition.
➔ Work on improving the model to
achieve high accuracy, aiming for over
95% accuracy on the test data. Handwritten Digits
WHAT IS CNN?
➔ A convolutional neural network (CNN) is a type of of artificial
neural network used in image recognition and processing
that is specifically designed to process pixel data.

➔ Input Layer: It represents the pixel matrix


of the image.
➔ Convolutional Layers: Apply filters
(kernel) across the image to extract
features.
➔ Pooling Layers: Reduce features map
size to save computing power and prevent
overfitting.
➔ Fully Connected Layers: Combine all
information to make the final prediction.
WHY CONVOLUTION?
➔ Convolution is a mathematical operation where a filter (or
kernel) slides over the input matrix, performs dot products to
create feature map that highlights key features.

➔ Convolutions provide better feature extraction.

➔ They save a lot of computation power as compared to ANNs.

➔ Convolutions require less parameters than fully connected


layers, reducing memory and overfitting.

➔ Due to less number of required parameters, lesser fully


connected layers are needed.
METHODOLOGY
Training Dataset
➔ Training of the network is done using the MNIST
dataset, which contains grayscale images of
handwritten digits (0-9).

➔ MNIST dataset has a training set of 60,000


examples, and a test set of 10,000 examples.

➔ All the images in the dataset are of 28x28 pixels.


Example of MNIST Dataset
CNN WORKING
➔ Input is Grayscale Image which is
represented in matrix with single
channel which is normalized to range
[0, 1].
➔ The filter/kernel (3x3 matrix) produces
a convolved feature by sliding over the
image, extracting patterns like edges.

➔ After convolution, activation function


(ReLU) adds non-linearity. Kernel View

➔ After convolution and activation, the feature maps


is flattened in 1D vector, and the fully connected
layer makes the final prediction.
MODEL ARCHITECTURE

➔ The model processes input images through convolutional


and pooling layers to extract features, flattens the feature
maps, and then classifies the digit using fully connected
layers.
RESULTS

Model Accuracy

Model Prediction
CONCLUSION
➔ Provided an in-depth overview of CNN basics, focusing
on its key layers.
➔ Showed how CNNs are well-suited for identifying patterns
in images, making them effective for handwritten digit
recognition.
➔ Developed a CNN model to recognize and classify
handwritten digits, achieving 97.7% accuracy on the
MNIST dataset.
➔ Presented a detailed overview of CNN architecture.
➔ Showcased the model’s predictive accuracy with sample
outputs.
REFERENCES
1. LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-Based Learning
Applied to Document Recognition. Proceedings of the IEEE, 86(11), 2278-2324.
2. Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). ImageNet Classification with
Deep Convolutional Neural Networks. Advances in Neural Information Processing
Systems, 25, 1097-1105.
3. Deng, L. (2012). The MNIST Database of Handwritten Digit Images for Machine Learning
Research. IEEE Signal Processing Magazine, 29(6), 141-142.
4. Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.
5. Zhang, Z., & Sabuncu, M. R. (2018). Generalized Cross Entropy Loss for Training
Deep Neural Networks with Noisy Labels. arXiv preprint arXiv:1805.07836.
THANK YOU

You might also like