The objective of this task is to build computer vision–based machine learning models to classify different celestial objects such as galaxies, stars, planets, nebulae, asteroids, and more using astronomical image data.
This is an open-ended task, and participants are encouraged to explore multiple approaches, architectures, and evaluation strategies.
The SpaceNet dataset is a curated collection of thousands of labeled astronomical images spanning multiple celestial object categories.
- Source: SpaceNet – Kaggle
- Suitability: Designed for deep learning–based image classification.
- Diversity: Balanced across multiple classes.
- Compatibility: Works with baseline CNNs and advanced architectures like Vision Transformers (ViT).
Important
Participants must download and attach the dataset themselves. If local resources are limited, use Kaggle Notebooks and upload the .ipynb files here.
To contribute, follow these steps:
- Fork the repository to your own GitHub account.
- Clone your fork locally or work directly on Kaggle Notebooks.
- Create a folder inside the
participants/directory named exactly as your enrollment number. - Add your notebooks (
.ipynb) inside your folder. - Commit your changes to your fork.
- Open a Pull Request to submit your work to the main repository.
participants/
└── <your_enrollment_number>/
├── notebook_1.ipynb
├── notebook_2.ipynb
└── ...
- Create a subfolder named exactly as your enrollment number
- Upload only your notebooks inside your folder
- Do not upload datasets, model weights, or large binaries
- Do not modify or delete other participants’ folders
- Any additional instructions, constraints, or evaluation criteria will be issue-specific
- Participants are expected to carefully read the relevant GitHub issue before making a submission
All doubts, clarifications, and discussions related to this task will be entertained via the Discord bot.
Please refrain from opening GitHub issues for general doubts unless explicitly instructed.