Dataset setup is unclear and causes errors when running Super Resolution pipeline

[Shrishagk]

Problem

While setting up the Super_Resolution_Atal_Gupta module, the dataset preparation process is not clearly defined. The current workflow leads to multiple issues for new users:

• No script is provided to generate HR–LR image pairs
• Running the pipeline without manually prepared data leads to FileNotFoundError
• The data loader in get_data_diff.py throws indexing errors when dataset size is small
• README lacks clear dataset generation instructions

This makes reproducing results difficult and increases onboarding time for contributors.

---

Observed Errors

1. FileNotFoundError due to missing dataset files
2. IndexError in get_data_diff.py when dataset size < required index
3. Confusion about required directory structure

---

Proposed Fix

I will submit a PR that:

✔ Adds a dataset generation script (generate_dataset.py)
✔ Updates README with step-by-step dataset setup
✔ Fixes indexing bug in get_data_diff.py
✔ Improves reproducibility of the pipeline

This will make the module runnable out-of-the-box and improve contributor experience.

[vignathi123-vi]

Hi I am GSoc 2026 applicant,

I’m interested in solving this issue and improving dataset reproducibility and onboarding. Is the issue open?

[Shrishagk]

Hi @vignathi123-vi ,Thank you for your interest in improving dataset reproducibility. The issue has been resolved through PR #136.

You’re welcome to review the changes and share any suggestions or improvements.

[vignathi123-vi]

Hi,
So while looking at both the problem and the solution, I just thought that training assumes perfectly aligned HR–LR pairs. A lightweight validation step (count, filename consistency) before training could prevent subtle errors and improve robustness. I also observed that dataset related choices (scale factor, patch size) are at present implicit but externalizing these into a configuration file would make experiments more transparent and reproducible without altering the training logic at the core.
Thank you!

[gaurav712-gujjar]

Hello 👋
I am a beginner contributor and interested in working on this issue.
I would like to work on improving the dataset setup documentation for the Super Resolution pipeline.

Can you please assign this issue to me?

Thank you!

[Shrishagk]

Hi @vignathi123-vi , Thank you for the suggestion. I agree that adding validation checks for HR–LR alignment and externalizing dataset parameters into a configuration file would improve robustness and reproducibility. I will work on incorporating these improvements.

[vignathi123-vi]

Hi @Shrishagk ,
It's totally my pleasure! I would love to incorporate these improvements in the project. Can I do so and submit the improved files? It would be great to receive your approval to start my work.
Thankyou!
Greetings,
Vignathi.