Welcome to the MIDOG 2025 Challenge! This page will guide you through the initial steps to get started with the competition. Here, you’ll find information on joining the challenge, accessing the datasets, setting up your environment, and exploring example workflows to help you begin developing your algorithms.
For more detailed information about the tasks of this years challenge, visit the Challenge Overview page.
Accessing the Data
All publicly available datasets are allowed to use for this years challenge. A list of available datasets can be found at the Datasets page. This page provides:
- A link to the paper explaining details about the datasets and annotations
- A link to the GitHub page with instructions how to download the data
Setting Up Your Environment
We provide an official GitHub repository with example implementations and utilities to help you get started. You can find it at MIDOG_2025_Guide. The repository includes:
- A notebook introducing the MIDOG++ dataset with exploratory analysis
- A notebook demonstrating the training of a simple object detection model
- A notebook introducing the MIDOG 2025 atypical dataset and providing an exemplary classification pipeline
- Utility functions to handle microscopy images
The provided notebooks are meant to serve as examples and starting points. We strongly encourage you to be creative and develop your own implementations to achieve good results.
Submission Guidelines
To participate in the challenge, you will need to submit your models and predictions using Docker containers. Detailed instructions for preparing and submitting your solutions will follow here!
Support and Community
We are here to help! If you have any questions or need assistance, you can join our Discord Community via this link and you will be able to connect with the other participants and the organizing team.
Feel free to reach out and share your insights, ask questions, or collaborate with others. We are excited to see what you will create!