Issue: #81 – Organize Dataset into Train/Test/Eval Splits

[Krishna200608]

Issue: #81

Summary

This Pull Request restructures the existing class-separated astronomy image dataset into
standard train, test, and eval splits to enable reproducible training and evaluation.

Work Done

• Implemented a reusable, class-based dataset splitting utility
• Organized images into train, test, and eval directories with a 70% / 20% / 10% split
• Preserved class-wise separation across all splits
• Verified split correctness by printing per-class image counts

Implementation Notes

• The dataset restructuring is performed locally within the Kaggle environment
• Symbolic links are used instead of copying files to avoid storage limitations
• No dataset files are uploaded to the repository; only the splitting logic is shared
• The implementation is added as a new section in the existing notebook under:
  participants/IIT2023139/data_exploration (version 2).ipynb

Reproducibility

• A fixed random seed is used to ensure deterministic splits
• The splitting logic is reusable and can be applied to similar class-structured datasets

Environment

• Platform: Kaggle Notebook
• OS: Windows 11 (local development)
• Editor: VS Code

This PR addresses only Issue #81 and does not overlap with any other issues.

[OpenGitBot]

Hey @Krishna200608

Thanks for opening this PR 🚀. Mentor will review your pull request soon and till then, keep contributing and stay calm.

Thanks for contributing in OpenCode'25 ✨✨!