Fix cuDNN runtime wiring for Paddle GPU

Enable GPU OCR in full HF app

Deploy updated core backend pipeline

Split OCR core from local studio and deploy HF core

Fix filename handling in HF demo; switch input to gr.File; ensure correct file reading and export to Excel

Add filename and title export to Excel; update hf_process to build DataFrame from title + sensors; improve file name handling

Fix ocr code

Fix ocr code

Fix ocr code

Add print of results

Add print of sensors

Fix ocr_sensors function

Fix ocr_sensors function

Add print for roi and ocr_results

Add print for roi and ocr_results

Fix pipeline code for OCR ensors

Fix OCR sensors: remove forced ROI height clamp (hh=17) causing unreadable ROIs

Update requirements: numpy, opencv, gradio, easyocr

Replace Tesseract with EasyOCR for title only

fix: correct color ranges

fix: correct color ranges parsing

Update demo pipeline

Final demo OCR version for Streamlit

Switch to PaddleOCR 2.6.1 for Streamlit compatibility

Add demo OCR version and Streamlit app

Fix: paddle_ocr parameters

Fix: add ocr_version

Fix: disable use_textline_orientation for Colab compatibility

Fix OCR output format, unpack rec_texts, round scores

Update project structure, add pipeline and tests