I honestly can't tell if you are being serious. Is there any doubt that the "OCR pipeline" will just be an LLM and it's just a matter of time? What you are des

I honestly can't tell if you are being serious. Is there any doubt that the "OCR pipeline" will just be an LLM and it's just a matter of time? What you are describing is similar to how computer used to detect cats. You first extract edges, texture and gradient. Then use a sliding window and run a classifier. Then you use NMS to merge the bounding boxes.