Giters

RudraxDave / AmericanSignLanguage_Reader

Project Data: The Dataset is 1.11 GB in Size. The images in the dataset are manually captured and not computer-generated. The dataset linked above contains images from 29 classes (26 alphabets, SPACE, DELETE, and NOTHING). Each class contains 3000 images in the training set and each image is a 200 x 200 RGB image. The training data set contains 87,000 images, of which 26 are for the letters A-Z and 3 classes for SPACE, DELETE, and NOTHING. These 3 classes are very helpful in real-time applications and classification. The test data set contains a mere 29 images, to encourage the use of real-world test images.>the test set is very small. We plan to Experiment with different architectures and hyperparameters. First, implement the CNN architecture described in the Models of [8] and apply our training data to generate the classification of 26 Letters. Our goal is to map images from a particular domain to a Letter that it means. We will be finetuning (transfer learning) existing models that perform classification on RGB images. We will Look at the levels of RGB images in Particular Centroidal Pixels and We can Classify from the Centre Which Shape is Defined. We are Planing to Use RESNET - Residual Neural Networks or Similar to be able to argue and analyze the better Accuracy between all of the Networks.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

EE541 Final Project - Spring 2022 - American Sign Language Reader using Deep Learning Algorithms

University of Southern California

Contributors

Rajath Ramegowda
Rudrax Dave

Note : Summarised Report is added in the current repository

Introduction

This project allows you to put your abilities learned in Introduction to Deep Learning in Python and Image Processing with Pytorch in Python into practice, such as creating convolutional neural networks to categorize photos. Computer vision systems that convert sign language to spoken language have made significant development in recent years. Complex neural network topologies are frequently used in this technology to identify tiny patterns in live video. However, understanding how to construct a translation system is the first step. In this notebook, we'll learn to categorize photos of American Sign Language (ASL) letters using a convolutional neural network. We will train the network and assess its performance after loading, inspecting, and preparing the data.

Goal

The goal of this project is to build a deep learning model to read the American Sign Language. Building 6 CNN models and comparing performance with predefined ResNet50 model.

Results

CNN 3 convolution layer with 89.99% accuracy ResNet50 with 100% accuracy

Note to access project files

CNN Models ipynb files

cnn_1Layer
cnn_3Layer
cnn_5Layer

ResNet50 ipynb file

ResNet50

Report

About

Project Data: The Dataset is 1.11 GB in Size. The images in the dataset are manually captured and not computer-generated. The dataset linked above contains images from 29 classes (26 alphabets, SPACE, DELETE, and NOTHING). Each class contains 3000 images in the training set and each image is a 200 x 200 RGB image. The training data set contains 87,000 images, of which 26 are for the letters A-Z and 3 classes for SPACE, DELETE, and NOTHING. These 3 classes are very helpful in real-time applications and classification. The test data set contains a mere 29 images, to encourage the use of real-world test images.>the test set is very small. We plan to Experiment with different architectures and hyperparameters. First, implement the CNN architecture described in the Models of [8] and apply our training data to generate the classification of 26 Letters. Our goal is to map images from a particular domain to a Letter that it means. We will be finetuning (transfer learning) existing models that perform classification on RGB images. We will Look at the levels of RGB images in Particular Centroidal Pixels and We can Classify from the Centre Which Shape is Defined. We are Planing to Use RESNET - Residual Neural Networks or Similar to be able to argue and analyze the better Accuracy between all of the Networks.

Languages

Language:Jupyter Notebook 100.0%