Back to Tracks

Duality AI Space StationHackathon

Duality AI

Train cutting-edge AI models for Object Detection in a space station environment using Duality AI's synthetic dataset from their digital twin simulation platform - Falcon.

Prize Pool: ₹17,000
🚀

Ready to Launch?

Join Duality AI's Falcon platform and start training your space station object detection model!

🛰️ Join Falcon Platform - Start Training

Challenge Overview

Space Station AI Training

Duality AI presents a challenge designed to test cutting-edge AI training techniques for Object Detection in a space station environment. Participants will train models using synthetic datasets from Duality AI's digital twin simulation platform, achieving accuracy and precision through parameter optimization.

AI Training

Learn industry-proven techniques to train AI models using only synthetic data

Digital Twins

Work with realistic virtual representations of space station environments

Object Detection

Detect and classify critical space station objects for operational safety

Challenge Objectives

Primary Goals

  • Train a robust object detection model using the provided synthetic dataset
  • Accurately detect and classify space station objects for operational safety
  • Evaluate model performance across challenging scenarios including varied lighting and occlusions
  • Benchmark and optimize for accuracy, generalizability, and efficiency

Target Objects

🧰 Toolbox

Essential maintenance equipment detection

🫁 Oxygen Tank

Critical life support system identification

🧯 Fire Extinguisher

Emergency safety equipment localization

Hackathon Structure

AI Engineering

  • • Train and fine-tune YOLOv8 model
  • • Evaluate model performance
  • • Optimize accuracy and inference time
  • • Use synthetic dataset from Falcon

Documentation

  • • Document complete workflow
  • • Create structured reports
  • • Prepare final presentations
  • • Include performance visualizations

Bonus Challenge

  • • Create application using the model
  • • Plan for model updates with Falcon
  • • Earn up to 15 bonus points
  • • Showcase real-world implementation

Judging Criteria

Model Performance

80 Points

mAP@0.5 (Mean Average Precision at IoU=50%)

Measures accuracy of object detection and classification. Higher mAP scores indicate better performance.

Report Clarity

20 Points

Structured Findings & Detailed Reporting

Well-organized documentation of methodology, challenges, and solutions with clear steps.

Bonus Points

+15 Points

Create an application using the detection model and describe how Falcon can keep it updated

Key Deliverables

Trained Model

  • • Fully trained YOLOv8 model
  • • Model weights and config files
  • • Training and inference scripts
  • • Three object category detection

Performance Report

  • • mAP@0.5 scores
  • • Confusion matrix analysis
  • • Failure case evaluation
  • • Performance visualizations

Documentation

  • • README with instructions
  • • Methodology documentation
  • • Environment requirements
  • • Reproduction guidelines
🚀

Ready to Launch?

Join Duality AI's Falcon platform and start training your space station object detection model with cutting-edge synthetic data!

🛰️ Join Falcon Platform - Start Training

Quick Setup

Environment setup with provided scripts

Quality Data

High-quality synthetic datasets from Falcon

Community Support

Discord support throughout the hackathon

← Back to All Tracks