Winter Projects GDSC IITK

“
The Winter Projects are designed to provide students with a
dynamic platform for learning and collaboration. Our dedicated
mentors, committed to guiding and supporting their mentees,
play a crucial role in ensuring the successful completion of these
innovative projects. We aspire to ignite the spark of creativity
and learning, fostering a collaborative and engaging
environment that empowers students to apply their skills to real-
world challenges.
2

Why apply for Winter Projects under GDSC IITK?
Beginner Friendly
Designed to be
exceptionally
beginner-friendly,
offering a supportive
environment tailored
for those taking their
first steps in the world
of technology
Theoretical
Expertise
Theory and materials
curated by the
experts at Google,
handcrafted to
provide a smooth
experience to one
and all
Hands-on
Application
offer a hands-on
approach, allowing
you to implement
your skills in real-
world scenarios with
practical experience
in terms of code
3

Why apply for Winter Projects under GDSC IITK?
Collaborative
Environment
Share ideas,
collaborate on
projects, and connect
with like-minded
peers who are equally
enthusiastic about
technology
Diverse Learning
Cover a wide
spectrum of
technology domains,
from machine
learning and app
development to web
development
Prepare for Future
Challenges
Winter Projects are
not just about the
current winter break;
they prepare you for
future challenges
(Solutions Challenge
from Google, Prize
money: $3000 per
person per team to
the top 3 teams)
4

List of Projects
Intro to Machine Learning Image Processing Active Learning
Chirp: Birdsong
Recognition
Web-Dev Basics and
Applications
Chat-Bot using RNNs
Forecasting using Time
Series Model
Intro to Generative AI App Dev Track
5

PROJECT SHOWCASE PORTAL
6
https://projects-gdsc-iitk.netlify.app/
List of all projects are here to check
out

Process of
Application
First Come First Serve
3 preferences offered
Your application response
matters (preferences and
enthusiasm)
7

Certification
All mentees who complete the
winter project successfully, will
be ratified by GDSC IITK
8

How many projects can you take?
You can officially take part in ONE project and will
receive ratification for only one project.
BUT
You can audit upto TWO OTHER projects and learn
more about your interests
9

Introduction to
Machine Learning
Project Mentors:
Kalika (8168084181)
Amay Raj (8696304685)
Expected Mentees Intake
20

Project Overview
Project Info
• Our project is designed to
immerse participants in the
core concepts of Machine
Learning (ML).
• Exploring Pandas for data
manipulation, Intermediate
ML techniques, and
Feature Engineering
methodologies.
• The project equips
participants with practical
skills for real-world ML
challenges.
12
Problem Statement
• Addressing the crucial
aspects of data
preprocessing, model
building, and feature
engineering.
• Empowering participants to
apply their acquired skills
to tackle complex ML
problems across diverse
domains.
Approach
• Fostering a hands-on
learning environment
leveraging a robust tech
stack, where participants
engage in practical coding
assignments and real-world
projects.
• Emphasis is placed on
applying theoretical
knowledge to real
scenarios, ensuring a
holistic understanding of
ML concepts.
Logistics
• The mentorship team
consists of Kalika (210482)
and Amay Raj (210116)
• Participants will have
access to curated tutorials,
hands-on coding exercises,
real-world projects, live
sessions, and a continuous
feedback mechanism.

Python
Pandas
Scikit-Learn
TensorFlow
Jupyter
Notebook
Google Colab
13
TECH STACK: Language, Libraries & Frameworks

Timeline & Learning Objectives
Week 1
Introduction to ML and Pandas, covering model workings, basic EDA, model validation, and Random Forests.
14
Week 2
Intermediate ML, focusing on missing values, categorical variables, pipelines, cross-validation, XGBoost, and data
leakage.
Week 3
Feature Engineering, exploring mutual information, creating features, K-means clustering, PDA, and target encoding.
Week 4
• Final Project
• Bonus: Surprise Project (subject to time limits).
Learning Objectives:
• Comprehensive understanding of data preprocessing, model building, and feature
engineering.
• Practical skills to tackle real-world ML problems.

Learning Resources
◎ Curated tutorials, articles,
documentation.
◎ Hands-on exercises using
Kaggle links for practical
reinforcement.
Interactive Sessions
◎ Live sessions for doubt
resolution and discussions.
Feedback Mechanism
◎ Regular feedback and
assessment for progress
tracking.
Projects
◎ Capstone project showcasing
participants' mastery of ML
concepts.
◎ Bonus Project if time permits.
15
Project Work Details

IMAGE PROCESSING
AND COMPUTER VISION
GDSC IITK Winter Project
PROJECT MENTORS:
PAVAN KUMAR C
AKSHAT MEHTA
ARCHISHA SINGH
25

OVERVIEW OF THE PROJECT
Understanding Image Processing
Techniques:
The project aims to familiarize
participants with the fundamental
concepts of working with digital
images. This includes comprehending
image representation, pixel
manipulation, color spaces, filters,
transformations, and various other
methods used to preprocess,
enhance, or modify images.
Integration with Machine Learning:
It involves connecting the knowledge
gained in image processing with
machine learning methodologies.
Participants will learn how to utilize
processed image data as input for ML
models. This integration can involve
tasks like image classification, object
detection, segmentation, and more,
where ML algorithms leverage
preprocessed image data for decision-
making.
"Unlocking the power of visual data through image processing and
machine learning illuminates a world where pixels tell stories, and
algorithms decode the language of images."
17

OVERVIEW OF THE PROJECT
Empowering Skills
for Visual Data
Handling:
The project intends to equip
individuals with the skills
necessary to work with
visual data. Images contain
a wealth of information, and
understanding how to
process, analyze, and
interpret this data is crucial
in numerous fields like
healthcare, autonomous
vehicles, security,
entertainment, and beyond.
Applications in
Computer Vision:
Understanding image
processing in the context
of ML leads to applications
in computer vision.
Participants will explore
how these techniques are
utilized in computer vision
tasks, such as recognizing
objects, detecting patterns,
understanding scenes,
and extracting meaningful
information from images.
Real-World Impact:
By gaining proficiency in
image processing
techniques integrated with
ML, participants can
contribute to various real-
world applications,
potentially enhancing
systems in medical
imaging, surveillance,
augmented reality,
autonomous vehicles,
quality control, and many
other domains where
visual data analysis plays
a critical role.
18

Python
OpenCV,
NumPy,
PIL, scikit-
learn
Jupyter
Notebook,
Google Colab
19
TECH STACK: Language, Libraries &Tools

20
1 2
4
3
Foundational
Image
Techniques
Advancing
Techniques and
ML Integration
Final Project:
Decoding with OCR
& CNNs
Empowering
Through Practical
Applications

Learning Resources
◎ Tailored tutorials, articles, and
documentation for image
processing and computer vision,
complemented by hands-on
exercises using Kaggle datasets
for practical reinforcement.
Interactive Sessions
◎ Live sessions for interactive
doubt resolution and discussions,
supplemented by curated
learning resources.
EVALUATION
◎ Every week, there will be two
sessions in which attendance is
necessary. Final project need to be
completed to get ratified for the
project completion.
Projects
◎ Capstone project showcasing
participants' mastery of ML
concepts.
◎ Bonus Project if time permits.
21
Project Work Details

Active Learning
GDSC IITK Winter Project
Shashwat Gupta
10

Introduction
The goal of this project is to familiarise the mentees with some Active Learning methodologies.
Potential use Case : Adaptive Tests, Labelling Diseases images
Recent Work : Deep Active Learning, CoreSets
Expectation:
1. Become comfortable with python and Machine Learning coding
2. Develop some cool stuff in Active Learning on simple datasets and baseline models
3. Getting the feel of diverse AL Strategies
23

Timeline and Workflow
24
Weekly Plans:
● Week 1: Installation, python basics and familiarity with Pytorch and libraries
● Week 2: Pytorch basics and implementing the baseline model for linear regression/classification
● Week 3: implementing simple AL strategies
● Week 4: implementing complex AL strategies
● Bonus/ Further: Further resources and application to complex problems, real world use cases
and ideas
Tentative workflow:
a. resources released - n-3
b. Introductory interaction - n
c. Doubt clearing interaction - n+3
d. Assignment Deadline - n+7

Tech Stack:
● Programming Languages: Python
● Libraries and Frameworks: Pytorch, basic ML libraries
like : numpy, pandas, scikit learn, and Matplotlib
● Tools: Jupyter notebook/ Google Colab/ VS Code
● Coursework : Elementary Probability and Statistics,
Linear Algebra (basic idea from school and first year
MTH should be good)
● Requirement : Enthusiasm and willingness to learn
25

What is Active Learning!
The active learner combines a model, a query
strategy and (optionally) a stopping criterion.
The overall objective is to minimize the
interactions between the oracle (annotator)
and the active learner.
Broad types:
1. Pool-based
2. Stream-based
3. Membership Query Synthesis
Uncertainty and Diversity
26

Chirp: Birdsong
Recognition
Project Mentors:
Sweta Kumari (201036)
Bhavna Jayswal (211213)
15

Problem Statement:
Birdsong recognition is crucial for ecological monitoring and biodiversity conservation. It helps
scientists and conservationists track bird populations, assess environmental health, and detect
changes in ecosystems. Birdcalls can vary significantly within and between species. Factors
such as environmental noise, overlapping calls, and individual variations further complicate
accurate recognition.
Approach:
The ’Chirp’ project aims to develop an efficient and accurate birdcall recognition system using
machine learning techniques. We will analyze patterns in our audio data (EDA), extract unique
features (FE) from it that help differentiate one birdcall from another, remove any background
noise and build a neural network model that will enable the identification of bird species based
on their unique vocalizations.
28

29
Tech Stacks
Programming Languages: Python
Libraries: NumPy, Pandas, Matplotlib, Seaborn, Librosa, Sklearn, PyTorch
Tools: Jupyter Notebook, Google Colaboratory
Pre-requisites
Basic understanding of Python programming.
Familiarity with Machine Learning concepts is beneficial and highly
recommended.

“
30
Week 1:
Familiarization with Python
and Exploratory Data Analysis
(EDA)
Week 2:
Feature Extraction (FE) and
Introduction to Application of
Machine Learning to Audio Data
Week 3 and 4:
Building the Neural Network
Model and Evaluating its
Performance
Timeline
Learning Objectives
✔ Cleaning and preprocessing text as well as audio data.
✔ Learning which essential audio features to extract and how to extract them.
✔ Building your own neural network model using PyTorch.

Evaluation Metrics
31
ATTENDANCE
PERFORMANCE EVALUATION OF MODEL
WEEK 2: FE ASSIGNMENT
WEEK 3: BASIC MODEL BUILDING
WEEK 1: EDA ASSIGNMENT
The final evaluation
Build your own model!
Extract features from a practice audio dataset
Exploratory Data Analysis of a practice CSV file (text data)
Minimum 80% attendance is mandatory (~12 total sessions)

Web Dev Basics and Applications
Mentors:
Saugat Kannojia
Hemant Choudhary
25

Overview of Project
This project aims to guide participants through the process of building a dynamic
and responsive web application using React for the frontend and Firebase for the
backend. Participants will learn the basics of HTML and CSS, delve into React for
web development, and explore Firebase for web functionalities, including
authentication, database management, and hosting.
Focus: HTML, CSS, ReactJS, Firebase (Authentication, Database, Hosting).
Approach towards the project:
Learn and apply concepts simultaneously from Week 1, integrating knowledge into
the project. Weeks 3-4 focus on practical application with collaborative coding and
iterative development.
33

Our process is easy
Learn
Apply
Iterate
and
Innovate
34

Tech Stack and Framework
● Languages: HTML, CSS, JavaScript
● Frontend Framework: ReactJS
● Backend: Firebase
● Frameworks for Advanced Features (Bonus): Cloud Functions, Cloud
Run
● Tools: GitHub for CI /CD
35

Project Timeline
36
● Week 1: Get started with the basics of HTML, Get started with the basics of CSS; Learn
Responsive Design (Optional)
● Week 2: Get started with React; Firebase for Web, Get started with Firebase: Basics, Doing More
with Firebase: Authentication, Database Management, Hosting; Integration with React and
Firebase
● Week 3 and Week 4: Project Portal Revamp, Hands on and adding several new features to the
project
○ Search support to be added
○ Filters to be added to search projects according to tags, date-range etc
○ Connecting to google sheets for fetching data of the projects
○ UI Revamp
○ CI/CD through Github Actions
○ More features TBA
● Bonus Resources: Core Web Vitals, AMP, Accessibility, SEO; Explore advanced Firebase
features like Cloud Functions and Cloud Run.

ChatBot using RNN
Mentor:
Shrey Mehta
Aryan Maurya
25

Project Overview
Problem Statement
● Objective: Responding to
queries from provided
stories using Natural
Language Processing and
Recurrent Neural Networks
● Relevance: Demonstrating
practical NLP application by
enabling context-based
question answering in the
chatbot
39
Approach
● Structured Learning:
Progression from basics
(PDF handling) to advanced
(RNNs) for comprehensive
understanding and
implementation.
● Practical Focus: Hands-on
exercises, real-world
projects, and interactive
sessions to reinforce
learning and skill application
Logistics
● Mentor:
Shrey Mehta
Aryan Maurya
● Resource Utilization:
Curated tutorials, exercises,
and interactive sessions for
learning reinforcement and
engagement tracking

Tech Stack
Recurrent Neural
Networks
Python
Jupyter Notebook
Google Colab
40

Timeline
Week 1
◎ Handling PDF with
python
◎ NLP Basics
Week 2
◎ Speech Tagging
◎ Text Classification
Week 3
◎ Sentiment Analysis
◎ Topic Modelling
42
Week 4
Assemble all the parts into
building a cool Chatbot
Exercises:
Practical exercises will be given every week for better
understanding
Work commitment: 12hrs/week

Forecasting using
Time Series model
Mentors
Sanyam Jain (sanyamjain21@iitk.ac.in)
Aditya Pandey (adityap21@iitk.ac.in)
15

Project Info
This initiative seeks to instruct mentees on diverse time
series models employed in forecasting across various
domains. Participants will gain proficiency in applying
appropriate time series models to datasets, such as stock
prices, to predict future trends. The program maintains a
balanced emphasis on both theoretical understanding and
practical implementation of these models.
44

Tech Stacks
● Programming languages: - Python
● Libraries and Frameworks: - Pandas,
Statsmodel (ARIMA and SARIMA),
Tensorflow
● Tools: - Jupyter Notebook, Google Colab
45

Timeline
46
WEEK 6
WEEK 5
WEEK 4
WEEK 3
WEEK 2
WEEK 1
a. Probability and
PDFS,PMFs
b. Normal
distribution
c. Expectation,
variance,covaria
nce.correlation
a. Hypothesis
testing
b. CLRM,OLS
estimators,
c. reading the
regression
table
a. Non
stationary AR
process
b. General
ARIMA
processes
a. Autoregressiv
e process of
Order 1
b. Correlograms
c. Stationarity
condition &
autocorrelatio
n, partial
correlation
functions,
a. Moving
Average
processes
(MA
processes)
b. ARMA
processes
a. SARIMA
processes
b. Final project

PROJECT DETAILS
47
Expected Duration
5-6 weeks
PREREQUISITE
Knowledge of MSO201/HSO201
is a plus but not necessary
Time Commitment
6-7 hours/week
MEETS
1-2 theory and doubt classes per
week

Evaluation
Every week, there will be two sessions in which
attendance is necessary. Final project need to be
completed to get ratified for the project completion.
48

Intro to
Generative AI
Mentors :
Anushka Gupta - 8319728058
Aniket Sharma - 9770289026
Paramveer Choudhary - 8529234309
20

PROJECT INFORMATION
Generative AI (Artificial Intelligence) refers to a subset of AI that focuses on creating data rather
than analyzing existing data. It involves algorithms that generate new, original content based on
patterns learned from a given dataset. Generative AI is crucial because it enables machines to
understand, learn, and replicate patterns, thereby producing new content, which is valuable in
various fields like art generation, text completion, and image synthesis. Its purpose lies in
enhancing creativity, aiding in content creation, and assisting in data augmentation for machine
learning models.
PROBLEM STATEMENT
50
This project aims to solve the challenge of imparting foundational knowledge of Generative AI and its
practical application to participants. The primary problem revolves around the complexity of
understanding key concepts like Large Language Models (LLMs) and Generative Adversarial Networks
(GANs) and applying this knowledge to create realistic numeral images from the MNIST dataset.

APPROACH AND LOGISTICS
We will proceed in the following steps.
1. Fundamentals of Generative AI: Introduction to the core principles and theories underpinning
Generative AI, providing participants with a solid understanding of its fundamentals.
2. Introduction to Large Language Models (LLMs): Familiarization with LLMs, highlighting their
significance in creating AI-generated content by comprehensively processing and
understanding language patterns.
3. Hands-on Experience with Generative Adversarial Networks (GANs): Practical engagement
with GANs, allowing participants to implement and experiment with these models that
consist of two neural networks, the generator and the discriminator, to generate new data
closely resembling the original dataset.
4. Application using MNIST Dataset: Utilizing the MNIST dataset as a practical case study,
enabling participants to build and train a generative model that can create realistic numeral
images (0-9).
We will provide the necessary resources, including access to programming environments with
essential libraries. Tasks will be assigned at each stage of the project to guide participants through
theoretical learning and practical implementation.
51

Tech Stack
Programming Languages : Python
Libraries and Frameworks: OpenCV, NumPy, Scikit-Learn, Tensorflow, Pytorch*
Tools: Google Colab
52

Timeline
53
Week - 1
Intro to generative
AI
And LLMs
Week -
2
Week -
3
Week -
4
Generative AI for
Developers
Generative
Adversarial Networks
(GANs)
Hands - On
Project

App Dev Track
Mentors :
Sahil Singh: sahilsingh20@iitk.ac.in
20

PROJECT INFORMATION
This project is designed to guide participants through the exploration and
hands-on learning of Flutter and Dart, the powerful frameworks for building
cross-platform mobile applications. Participants will start with the installation
of the necessary tools, explore the basics of Dart and Flutter, build their first
Flutter app, and delve into advanced topics such as Material Design,
networking, state management, using plugins, and testing.
55

APPROACH AND LOGISTICS
Development:
● Set up the development environment: Ensure you have the necessary tools and libraries installed,
including Flutter SDK, DartPad, and other relevant packages.
● Implement the core functionalities: This includes building the user interface using Flutter widgets,
implementing state management techniques, and integrating networking features if necessary.
● Focus on code quality: Write clean, modular, and well-documented code. Utilize testing practices to
ensure the application is stable and bug-free.
● Utilize plugins and packages: Leverage the vast ecosystem of Flutter plugins and packages to
enhance your application with additional features and functionalities.
This structured project plan ensures a step-by-step exploration of Flutter and Dart, covering installation, basic
concepts, app building, advanced topics, and practical application through real-world projects. Feel free to
customize this template based on your specific project details and requirements.
56

Tech Stack
Programming Languages : Dart
Libraries and Frameworks: Flutter and Pub.dev packages
Tools: Android Studio
57

Timeline
58
WEEK 5
WEEK 4
WEEK 3
WEEK 2
WEEK 1
a. Installation
b. Explore Flutter
c. Building your first app
a. Adding Material Design
b. Networking
c. State Management
d. Plugins for Native APIs
e. Testing your code
a. Project Building: Developing a useful
application (more details on project later)

Application Link
https://forms.gle/2PC5G6
wrRhA8RpPN9
DEADLINE:
9th December, 11:59pm
59

Thanks!
Any questions?
You can find us at:
gdsc.iitk@gmail.com
60
Join us

Winter Projects GDSC IITK

More Related Content

Similar to Winter Projects GDSC IITK (20)

Recently uploaded (20)

Winter Projects GDSC IITK