feat: support multiple OCR parameter sets and improve text selection accuracy

[tvone]

Changes

• Added support for multiple OCR parameter sets (ocr_params_list) instead of a single config.
• Improved detection accuracy by testing various configurations (e.g., Det.limit_side_len min/max) to reduce missed detections of short or long text lines within images.
• Multiprocessing support (multiprocessing.Pool) can optionally be used to speed up processing by running different parameter sets in parallel.

Notes

• Default behavior remains the same if ocr_params_list is not provided.
• Parallel execution is optional and only applied if explicitly implemented by the user.

[Copilot]

Pull Request Overview

Enhance OCRProcessor to accept multiple OCR configurations and select the most comprehensive text result per image, adjusting the entry point to use the new ocr_params_list.

• Refactored OCRProcessor.__init__ and get_ocr_result to handle a list of parameter sets.
• Updated single_rec to iterate over all engines and pick the longest text output.
• Changed RapidVideOCRInput and RapidVideOCR in main.py to use ocr_params_list.

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
rapid_videocr/ocr_processor.py	Support multiple OCR parameter sets in single_rec, updated engine initialization and result logic.
rapid_videocr/main.py	Replaced ocr_params with ocr_params_list in input schema and processor instantiation.

Comments suppressed due to low confidence (1)

rapid_videocr/ocr_processor.py:61

• Add unit tests to verify that single_rec selects the longest text among multiple OCR parameter sets, covering this conditional branch.

                if max_txt_len < len(txts):

[SWHL]

Thanks, and I will merge it later.

[tvone]

Thanks, and I will merge it later.

Thank you.

[SWHL]

I'm really sorry for the late response; I've only just found the time to look into RapidVideOCR recently. I've carefully reviewed your submission. The core idea is to instantiate multiple OCR instances with different configurations simultaneously, perform multiple recognitions on the same image, and select the longest result as the final output.

However, this approach would lead to increased resource consumption. Out of curiosity, could you please share what specific scenario prompted you to consider this solution?