DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING

INTERNSHIP PRESENTATION
On
“Artificial Intelligence and Data Science”
Carried out at
Abeyaantrix Edusoft, Davanagere.
during
5th February 2025 – 15th May 2025

Presented by: External Guide: Internal Guide:

Chetan S K Ms. Namitha Banu Dr.. Pradeep T M
4GM21EC023 Internship Trainer Assistant Professor
Abeyaantrix Edusoft, Davanagere
Contents
• About the Company/organization where internship was carried out
• Area of Internship
• Details of Software and hardware knowledge acquired
• Introduction of the project undertaken
• Aim of project undertaken
• Theoretical background
• Design and Implementation details
• Flowchart
• Results & Discussion
• Conclusion (Inference drawn)
• Outcome of the internship
• Certificate (display the certificate provided by the organization)
• References
15/05/2025 Dept. of ECE., GMIT, Davangere 2
 About the Company
Company Name: Abeyaantrix Softlab (OPC) Pvt. Ltd.

Location: Vidyanagar, Davanagere, Karnataka – 577005

Website: www.asoftlab.com

Industry: Information Technology & Software Training

Overview:
• The internship was carried out as part of the academic curriculum in the field of Artificial
Intelligence and Data Science, with the aim of providing practical exposure to real-world
applications of theoretical concepts learned in the classroom. The internship took place at
[Company/Organization Name], an organization engaged in innovative solutions and research
in AI and data-driven technologies.
• During the internship period, I was actively involved in various tasks that included data
preprocessing, model development, evaluation, and deployment. The training focused on using
modern tools and frameworks like Python, TensorFlow, Scikit-learn, and Jupyter Notebook, as
well as understanding the importance of clean data, model accuracy, and performance
optimization.
Dept. of ECE., GMIT, Davangere
06/30/2025 3
 Area of Internship

Domain:
Artificial Intelligence and Data Science
Description:
During the internship, I was assigned to work on tasks related to:
• Data Collection and Preprocessing: Cleaning raw datasets, handling missing
values, performing feature selection and engineering using Python libraries
such as Pandas and NumPy.

• Exploratory Data Analysis (EDA): Understanding data distributions and

patterns using Matplotlib, Seaborn, and other visualization tools.

• Machine Learning: Building and evaluating models using algorithms

• Deep Learning: Implementing and training neural networks using

TensorFlow and Keras for tasks such as image classification .
Dept. of ECE., GMIT, Davangere
06/30/2025 4
 Details of Software knowledge acquired
During the course of the internship, I acquired hands-on experience and practical
knowledge in a variety of software tools, programming languages, libraries, and
platforms that are essential in the fields of Artificial Intelligence and Data Science. The
following outlines the key software skills gained:

1. Programming Languages and IDEs

• Python: Primary language used for data analysis, machine learning, and deep learning
tasks.

• Jupyter Notebook: Used extensively for writing and testing Python code in an
interactive environment.

• Google Colab: Leveraged for cloud-based Python coding and GPU-accelerated

training.

2. Data Analysis and Manipulation

• NumPy: For numerical computing and array manipulations.

• Pandas: For data preprocessing, handling missing values, and working with
dataframes.

Dept. of ECE., GMIT, Davangere

06/30/2025 5
3. Data Visualization
• Matplotlib & Seaborn: For creating plots, graphs, and visualizing datasets.
• Plotly: For building interactive visualizations and dashboards.
4. Machine Learning Libraries
• Scikit-learn: For implementing classical machine learning algorithms like
regression, classification, and clustering.
• XGBoost: Used for gradient boosting tasks and improving model performance.
5. Deep Learning Frameworks
• TensorFlow & Keras: Used to build and train neural networks for tasks like image
classification and natural language processing.
6. Natural Language Processing (NLP) Tools
• Hugging Face Transformers: For implementing advanced NLP models like BERT
and GPT-based architectures.

Dept. of ECE., GMIT, Davangere

06/30/2025 6
 Introduction of the project

• Sign language is a vital mode of communication for the hearing-impaired community.

• Recognizing hand-signed digits can aid in bridging communication gaps using

technology.

• Image-based gesture recognition is a key step in automating sign language translation.

• Convolutional Neural Networks (CNNs) are highly effective for visual pattern
recognition tasks.

• This project aims to accurately identify sign language digits using a CNN-
based model

Dept. of ECE., GMIT, Davangere

06/30/2025 7
 Aim of projects
• To develop a system that accurately recognizes hand gestures representing digits in
sign language.

• To apply computer vision and deep learning techniques for gesture classification.

• To create an assistive tool that bridges communication between deaf individuals and
the hearing community.

• To enable real-time recognition of sign language digits using image or video input.

Dept. of ECE., GMIT, Davangere

06/30/2025 8
 Theoretical Background
1. Sign Language digit Recognition : A computer vision task focused on identifying
hand
gestures that represent letters, digits, or words in sign languages such as ASL (American
Sign Language).
2.Computer Vision: A field of AI that enables machines to interpret and understand
visual
information from images or videos, crucial for recognizing hand shapes and positions.
3.Convolutional Neural Networks (CNNs): A deep learning architecture particularly
effective for image classification tasks; used here to extract and learn spatial features
from hand sign images.
4.Image Preprocessing: Techniques such as resizing, normalization, and augmentation
are applied to input images to improve model accuracy and generalization.
5.Gesture Recognition: A broader field involving the interpretation of human gestures
via mathematical algorithms, of which sign digit recognition is a specific application
focused on hand signs.
Dept. of ECE., GMIT, Davangere
06/30/2025 9
 Design and Implementation
1. System Design Overview:
• The trained model predicts digits from hand sign images in real-time
for output display.
2. Data Collection:
• The dataset consists of labeled images of hand signs representing digits 0–9, used to
train and evaluate the CNN model.
3. Data Preprocessing:
• Resized all images to a fixed dimension (e.g., 48x48 pixels) to maintain uniformity.
• Converted images to grayscale to reduce computational complexity.
• Normalized pixel values to a range between 0 and 1 for faster convergence during
training.
• Applied data augmentation techniques such as rotation, zoom, and flipping to increase
dataset variability and prevent overfitting.

Dept. of ECE., GMIT, Davangere

06/30/2025 10
4. CNN Model Architecture:

CNNs are great for image recognition tasks.

The structure usually includes:
• Convolutional Layers (Conv2D): These detect features (edges, shapes) in the image.
• Activation Function (ReLU): Applies non-linearity so the model can learn complex patterns.
• Pooling Layers (MaxPooling2D): Downsample the image to reduce computation and focus on
important features.
• Flatten Layer: Converts 2D features into a 1D vector.
• Dense Layers (Dense): Fully connected layers to make predictions.
• Dropout Layer: Prevents overfitting by randomly disabling neurons during training.
• Output Layer: Uses softmax to output a probability for each digit (0–9).

06/30/2025 11
5. Model Training:
• Feed the training data to the model.
• The model adjusts its internal weights to minimize the difference between its
predictions and the true labels.
• Use a loss function like categorical_crossentropy and an optimizer like adam.
6. Model Evaluation:
• After training, test the model on unseen images.
• Measure its accuracy — how many images it correctly classifies.
7. Programming Language: Python
• Libraries: TensorFlow, Keras, OpenCV, NumPy
• Development Environment: Jupyter Notebook, Google Colab

Dept. of ECE., GMIT, Davangere

06/30/2025 12
 Flowchart

Dept. of ECE., GMIT, Davangere

06/30/2025 13
 Results & Discussion
The CNN model achieved high accuracy in classifying sign language digits,
demonstrating effective feature extraction from hand gesture images.

Training and validation accuracy showed minimal overfitting, indicating that the model
generalized well to unseen data.

Confusion matrix analysis revealed occasional misclassifications between similar

gestures, highlighting potential improvements through data augmentation or deeper
architectures.

The model outperformed traditional machine learning approaches, affirming CNN’s

suitability for visual recognition tasks like sign language digit classification.
Dept. of ECE., GMIT, Davangere
06/30/2025 14
Precision and recall metrics confirmed consistent performance across all digit classes,
with slight variance in digits involving complex finger positions.

Real-time testing validated the model's responsiveness and accuracy, making it suitable
for integration into assistive communication tools.

06/30/2025 Dept 15
 Conclusion
• The project successfully demonstrates the use of deep learning for recognizing sign
language digits with high accuracy.
• It provides an effective and accessible tool for bridging communication gaps between
hearing and deaf individuals.
• The use of CNNs ensures accurate feature extraction and classification of complex
hand gestures.
• Real-time implementation shows the potential for practical applications in assistive
technology.
• Future improvements can include recognizing full sign language alphabets and
integrating multilingual support.

Dept. of ECE., GMIT, Davangere

06/30/2025 16
 Outcome of the internship
• Successfully trained a CNN-based model to recognize sign digit with
up to 70%+ validation accuracy on the MNIST-like Sign language
digits dataset.
• Identified limitations and proposed improvements for better model
generalization.
• Understood how to deal with imbalanced datasets and the importance
of data augmentation.
• Gained the ability to diagnose model performance through visual tools
like learning curves and confusion matrices.
• Applied logical reasoning to debug model issues and optimize
performance.

Dept. of ECE., GMIT, Davangere

06/30/2025 17
 Dept. of ECE., GMIT, Davangere
06/30/2025 18
 References
• https://www.geeksforgeeks.org/data-visualization-using-matplotlib/
• https://www.guru99.com/download-install-tableau.html
• https://jupyter.org/
• https://www.tableau.com/support/releases
• https://www.pcmag.com/reviews/tableau-desktop

Dept. of ECE., GMIT, Davangere

06/30/2025 19