`

HACKATHON
Problem Statements

ARTIFICIAL INTELLIGENCE

Theme: AI for Business Intelligence & Automation

The Code Conjurer - Natural Language to Code Snippets

Objective:
Build an AI-powered assistant that converts plain English descriptions into
executable code snippets in multiple programming languages.

Features:

 User-friendly UI with a prompt box for natural language input.

 Language selection dropdown (Python, JavaScript, Java, etc.).
 Display generated code with syntax highlighting.
 Export code or copy to clipboard.
Key Parameters:

 Input: Natural language description

 Output: Code snippet (Python, JS, Java, etc.)
 Language Selection: Dropdown or input selection
 Model Used: Transformer (Codex, CodeT5)
 Validation: Syntax checker, linter
 Interface: Text input box, code output area

1 | www.atomm.in
Any form of reproduction, dissemination, copying, disclosure, modification, distribution and or publication of this material
is strictly prohibited.
Theme: AI for Creativity & Entertainment

The Style Weaver - Artistic Style Transfer

Objective:
Extract fashion trends and visual elements from social media using AI and
apply style transfer for fashion branding.

Components:

 Data Mining: Use scraping tools to gather Instagram posts.

 Image Analysis: Use CNNs to identify colors, textures, styles.
 Text Analysis: NLP for captions, hashtags, emojis.
 Generative AI: Use GANs or VQGAN+CLIP to generate new fashion
visuals.
Key Parameters:

 Input: Social media image + text (captions, hashtags)

 Output: Stylized image or visual trend analysis
 Image Processing: CNNs for color/style extraction
 Text Processing: NLP (hashtags, sentiment, keywords)
 Generative Models: GANs, VQGAN+CLIP
 Clustering: K-means, DBSCAN for fashion categories

Theme: AI Animation and Visualizations

The Motion Painter - AI for Animating Still Images

Objective:
Animate static images by adding realistic or stylistic motion effects.

Features:

 Upload image interface.

 Select motion zones (e.g., water, trees).

2 | www.atomm.in
Any form of reproduction, dissemination, copying, disclosure, modification, distribution and or publication of this material
is strictly prohibited.
  Apply pre-trained models for motion synthesis.
 Export as video or GIF.
Key Parameters:

 Input: Still image

 Motion Selection: User-defined or auto-detected areas
 Output: Short animation (GIF/MP4)
 Motion Techniques: Optical flow, video GANs
 UI Tools: Drawing tool to mark animated regions
 Export Formats: MP4, GIF, WebP

The VFX Assistant - Simple Visual Effects Automation

Objective:
Simplify visual effects editing using AI automation.

Features:

 Background removal without green screen.

 Object tracking for overlays.
 Basic rotoscoping (object isolation).
Key Parameters:

 Input: Video or image

 Output: Modified visuals (removed background, tracked object)
 Segmentation: DeepLabV3, U-Net
 Tracking: DeepSORT, OpenCV trackers
 Tools: Background removal, rotoscoping, motion overlay
 Object tracking (OpenCV, DeepSORT).

3 | www.atomm.in
Any form of reproduction, dissemination, copying, disclosure, modification, distribution and or publication of this material
is strictly prohibited.
The Visual Explainer - Live Stream Object/Concept Recognition

Objective:
AI system that identifies key concepts/objects during live video streams.

Features:

 Real-time video analysis.

 Pop-up explanations or tooltips. Vocabulary builder for students.
Key Parameters:

 Input: Live video stream

 Output: Labels or tooltips with identified concepts
 Detection: YOLOv8, EfficientDet
 Concept Mapping: Pre-defined ontology
 Display: Real-time overlay with explanations

Theme: AI in Cybersecurity

The Phishing Detective - Email Fraud Detection

Objective:
Detect and flag malicious emails to prevent phishing attacks.

Features:

 Scan subject lines, body text, links, and metadata.

 Assign risk scores.
 Visual warning dashboard for inboxes.
Key Parameters:

 Input: Email content, headers, metadata

 Output: Risk score, threat label (Phishing/Not)
 NLP Techniques: BERT for classification

4 | www.atomm.in
Any form of reproduction, dissemination, copying, disclosure, modification, distribution and or publication of this material
is strictly prohibited.
  Features: URL check, keyword spotting, header anomalies
 Dashboard: Risk visualization, email metadata
DATASETS:

The Code Conjurer – Natural Language to Code Snippets

CMU CoNaLa Dataset: Contains 2,379 training and 500 test examples of
natural language intents paired with Python code snippets. Papers with
Code+1Papers with Code+1

TinyCodes Dataset: Offers 1.6 million short code snippets across various
programming languages, aiding in training models for code generation.
Hugging Face

The Stack: A comprehensive dataset of source code from GitHub

repositories, useful for pre-training code language models. Hugging Face

Theme: AI for Creativity & Entertainment

The Style Weaver – Artistic Style Transfer

Stylized Image Dataset: Ideal for training neural style transfer models, this
dataset includes images with various artistic styles. Kaggle+1GitHub+1

Anime Style Transfer Dataset: Comprises 820 training and 93 test pairs of
real faces and their anime-style representations, suitable for style transfer
tasks. Papers with Code+1GTS AI+1

Theme: AI in Cybersecurity

The Phishing Detective – AI for Identifying Malicious Emails

Phishing Email Dataset (Kaggle): Comprises approximately 82,500 emails,

with a balanced mix of spam and legitimate messages, aiding in training
phishing detection models. ISU Cose+3Kaggle+3UT Research Info+3

Phishing Email Detection v2.0: Offers a comprehensive dataset combining

email messages and URLs for phishing detection tasks. Hugging Face

5 | www.atomm.in
Any form of reproduction, dissemination, copying, disclosure, modification, distribution and or publication of this material
is strictly prohibited.
FULL STACK DEVELOPMENT

Theme: The Future of Work & Learning

The Team Catalyst - Project Management Tool

Objective:
Enable better team collaboration through smart task tracking.

Key Parameters:

 Modules: Project creation, task assignment, chat, deadlines

 Users: Admins, team members
 Statuses: To-Do, In Progress, Done
 Timeline: Gantt chart, calendar
 Backend: API for CRUD operations
 Realtime: WebSockets or Firebase
Features:

 Create/manage projects.
 Assign tasks, set deadlines.
 Kanban boards, Gantt charts.
 Real-time chat and notifications.

The Learning Path - Personalized Learning Recommender

Objective:
Provide personalized learning paths using user profiles and goals.

Key Parameters:

 Input: User interests, skill level, goal

 Output: Ranked list of resources
 Recommendation Engine: Collaborative filtering, NLP ranking

6 | www.atomm.in
Any form of reproduction, dissemination, copying, disclosure, modification, distribution and or publication of this material
is strictly prohibited.
  Content Types: Video, courses, articles
 Feedback: Rating system to improve future suggestions
Features:

 User onboarding to set goals.

 Recommender engine for courses/videos/articles.
 Feedback system for improved suggestions.

Theme: Facilitating Travel & Exploration

The Local Gem Finder - Hyperlocal Discovery

Objective:
Help users discover underrated locations and services based on
hyperlocal data.

Features:

 Location-based recommendations.
 Community ratings and tips.
 Map-based interface.
Key Parameters:

 Input: User location, interests

 Output: List/map of local spots
 Data Sources: Crowd-sourced reviews, geo-tagging
 AI Techniques: Clustering, recommendation systems
 Interface: Map UI with search/filter

Theme: Building Better Infrastructure & Services

The City Reporter - Community Issue Reporting

Objective:
Platform for citizens to report local issues and track resolutions.
7 | www.atomm.in
Any form of reproduction, dissemination, copying, disclosure, modification, distribution and or publication of this material
is strictly prohibited.
Features:

 Upload image and location tag.

 Issue classification (e.g., pothole, lighting).
 Notify relevant municipal body.
Key Parameters:

 Input: Image, description, location

 Output: Issue ticket with tracking ID
 Classification: NLP categorization (pothole, light, etc.)
 Location Tagging: Google Maps API
 Authorities: Notification via email/API

Theme: Creating Engaging Digital Experiences

The Interactive Narrative - Branching Story Platform

Objective:
Allow creators to write interactive stories and readers to explore narrative
paths.

Features:

 Story builder with branching logic.

 Save user progress.
 Comment and share system.
Key Parameters:

 Story Structure: Tree/graph of decisions

 User Input: Choices that affect next story part
 Author Tools: Drag-and-drop story map builder
 Reader Features: Save progress, bookmarks

8 | www.atomm.in
Any form of reproduction, dissemination, copying, disclosure, modification, distribution and or publication of this material
is strictly prohibited.
  Format: JSON for stories, HTML5/CSS for front-end rendering

9 | www.atomm.in
Any form of reproduction, dissemination, copying, disclosure, modification, distribution and or publication of this material
is strictly prohibited.