Travvir/panoocr — reverse-engineered prompt

Build me a Python library that can read text from 360 degree panorama images, not just normal flat photos.

I want to give it an equirectangular panorama file and choose an OCR engine, like Apple Vision on Mac, RapidOCR, EasyOCR, PaddleOCR, Google Vision, Gemini, or Florence 2. The library should automatically split the panorama into useful perspective views, run OCR on those views, remove duplicate text found in overlapping areas, and return the final text with confidence plus yaw and pitch positions on the sphere.

Please include a simple Python API where I can create an engine, run recognition on an image, loop through the results, and save them as JSON. Also support multi scale recognition for small text, adjustable perspective settings, and deduplication options.

Add a lightweight browser preview tool where I can drag in the panorama and JSON result file and see the detected text placed back onto the 360 image. Include examples and clear install instructions for the different OCR backends.