About the challenge

Foundation models are rapidly changing the way we approach machine learning tasks. Are you ready to dive into the future of machine learning? This is your chance to make a real impact! In our ML4Earth Foundation Model Hackathon, you'll unleash the power of cutting-edge foundation models to create high-resolution maps from aerial imagery.

What’s in Store?

  • Choose Your Weapon: Select and wield a state-of-the-art foundation model.
  • Hands-On Experience: Fine-tune your model with the datasets provided by us.
  • Challenge Your Limits: Test your skills and push the boundaries of what's possible.

Get Started

You can find more information and a Starter Pack at https://github.com/zhu-xlab/ML4Earth-Hackathon-2024

Requirements

What to Build

In this Hackathon, your mission is to harness the power of foundation models to produce high-resolution, detailed maps from aerial imagery. Here's what you'll need to focus on:

  1. High-Resolution Mapping:
    • Goal: Create detailed, accurate maps from aerial imagery.
    • Scope: Focus on urban planning, environmental monitoring, and other real-world applications where high-resolution mapping can make a difference.
  2. Model Selection & Fine-Tuning:
    • Choice: Select a foundation model that best fits the task—whether it’s a transformer, CNN, or any other advanced architecture.
    • Customization: Fine-tune your chosen model using the curated datasets we provide to enhance its performance for your specific mapping task.
  3. Data Integration:
    • Training Data: Utilize our comprehensive training dataset to teach your model.
    • Testing Data: Validate your model's accuracy and efficiency with our testing dataset.
  4. Impact & Usability:
    • Real-World Applications: Think about how your project can be applied in real-world scenarios. Highlight its potential impact on industries like agriculture, urban development, disaster response, and more.
    • Scalability: Ensure your solution is scalable and can handle large datasets efficiently.

What to Submit

  • Performance Metrics: An analysis of your model’s performance, including accuracy, speed, and any other relevant metrics.
  • High-Resolution Maps: The final output maps generated by your model.
  • Demo/Pitch Video: A short video demonstrating your project's capabilities and showcasing its features.
  • Source Code: A repository containing all the code used in your project, with clear instructions on how to run it.

Hackathon Sponsors

Prizes

1,800 in prizes
1st Prize
1 winner

600€ in cash paid by our sponsor, synoptic.com
400€ in books, GPU resources, and journal subscriptions.

2nd Prize
1 winner

300€ in cash paid by our sponsor, synoptic.com
200€ in books, GPU resources, and journal subscriptions.

3rd Prize
1 winner

100€ in cash paid by our sponsor, synoptic.com
200€ in books, GPU resources, and journal subscriptions.

Devpost Achievements

Submitting to this hackathon could earn you:

Judges

Prof. Xiao Xiang Zhu

Prof. Xiao Xiang Zhu
TUM

Georgiy Nefedov

Georgiy Nefedov
Synoptic

Dr. Adam Stewart

Dr. Adam Stewart
TUM

Yi Wang

Yi Wang
TUM

Judging Criteria

  • Model Performance
    How accurate are your model's predictions?
  • Computational Efficiency
    How efficient is the model in terms of computational resources?
  • Innovation
    Creativity in adapting the foundation model to the given task
  • Final-Pitch Performance
    Quality of the presentation of the results

Questions? Email the hackathon manager

Tell your friends

Hackathon sponsors

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.