Questions and minor concerns about `utils.py` design choices

[VibezCoder]

This issue is about github.com/humanai-foundation/RenAIssance/blob/main/RenAIssance_CRNN_OCR_Shashank_Shekhar_Singh/utility/utils.py.

1. Suggestions / potential issues

• Add error checking for cv2.imread(...) and pdf_to_images(pdf_path, output_folder) to handle missing/invalid files.
• remove_punctuation() (using string.punctuation and str.translate) seems unused/commented out; should it be removed or wired into the pipeline?
• In create_csv_from_folder, the check if file_name.lower() == ".png": continue looks unnecessary / possibly incorrect, since it checks whether .png exists, not .contains().
• Several magic numbers (e.g. last_textfile_number = 7, vertical_distance_between_lines = 10, count_semicolon - 2) aren’t documented; are these dataset-specific or heuristics that should be configurable?

2. Clarification questions

• In split_and_save_image, there’s no handling for the “in‑between” case 350 <= width <= 450; are such images guaranteed not to exist, or should they be treated as single/dual pages?
• The bounding box coordinate unpacking repeats variables (x_min, y_min, x_max, y_min, x_max, y_max, x_min, y_max); is the input format really repeated points, or is this an oversight?
• The augmentation helpers (rotation_aug, gaussian_noise_aug) write outputs into the same folder without checks; is the expected workflow to run them only once on a clean folder, or would it make sense to separate input/output directories?

I’m happy to open follow-up PRs to adjust these if you confirm which parts are intentional vs accidental.

Thank you

[tushar152006]

Hi, I would like to start by addressing the missing error handling for cv2.imread() and pdf_to_images() to properly handle missing or invalid files.
I'll open a focused PR covering this first so we can loop step wise.
Please let me know if you'd prefer a different approach

[Deepthanush1337]

Nice catch on these. I was looking through utils.py and noticed the same things.

The file_name check on line with if file_name.lower() == ".png": continue is definitely wrong - it's comparing the entire filename to the string .png instead of checking if it ends with .png. Should probably be ile_name.lower().endswith(".png") or just not needed at all since the function already filters by extension.

Also the bounding box unpacking with repeated variable names looks like a copy-paste issue to me - x_min and y_min get reassigned halfway through which would mess up the coordinates.

Since @tushar152006 is already handling the error checking part, I'd be happy to take on fixing the filename check and the bounding box variable issue separately if that works.

[Apprentice2907]

Hi,

Since the error handling and filename/bounding box fixes are already being worked on, I’d like to take a look at the remaining concerns, particularly the unused remove_punctuation() function and the undocumented magic numbers.

If that sounds good, I can work on a small PR addressing those.

Thanks!