OCR Test 6 was easier: The image contained the number in horizontal orientation and with no disturbing background. Google's reCAPTCHA service with house numbers. After all, Google uses house numbers in its reCAPTCHA service. Number recognition inside an image is not easy. Amazing! In other words: XKCD speech bubbles are so difficult to decode, they could be used as CAPTCHAs :) OCR Test 6: Recognize a number in an image Surprise #4: While OnlineOCR received a passing grade with its “poor” result, all other services failed completely. XKCD comics are fun – and not difficult to read for humans (just sometimes difficult to understand?). OCR Test 5: Recognize an XKCD comic (tricky font) No longer a surprise: Google OCR did again worse then the free OnlineOCR service and Abbyy’s commercial OCR solution. None of the OCR engines did well in this benchmark, plenty of room for improvement. Test 4: Mobile Phone Image OCR OCR Service Usually you are not in a photo studio when archiving your creased receipts or coffee-stained invoices. The result is a creased and slightly wrinkled printout. This makes the task more realistic in a reproducible way. For this test the meanwhile well-known NY times article was printed out, and then folded/unfolded. Your camera as a document scanner? That is possible in 2015. After all, Google is the company that builds self-driving cars and won the ImageNet Visual Recognition Challenge last year. Surprise #2: Google OCR did worse than the “noname” OnlineOcr service (and Abbyy)! My clear expectation at the start of this market research was that Google’s OCR service will set the gold standard. Nevertheless the image still easily readable for humans. Test setup: The printout was scanned at a resolution of 75 dpi, printed out again and scanned a second time at 75dpi. No big surprises here, OCR on high-quality scans worked ok with all services. Test 2: High-quality Scan OCR OCR Service As some services do not take PDF format as input, the JPEG (JPG extension) format is used as the lowest common denominator in all tests. Test setup: A printout of the NY Times article was scanned at a resolution of 100dpi. With PDF OCR X, a desktop OCR software that uses the Tesseract engine. This image shows the verification of the OCR result If someone can explain the bad result, I would be very interested to hear! Disappoint results with Tesseract all over. The result is not as bad as in the Tesseract online demo, but still poor. I doubled checked the result with PDF OCR X, a Windows/Mac tool that wraps the Tesseract-OCR engine. According to Google “Tesseract is probably the most accurate open source OCR engine available.” This makes me feel a bit like an elementary school teacher who has to give bad grades to the son of the major - so I first blamed the online version of Tesseract (at CustomOCR) that I am using. Surprise #1: Yes, Tesseract flunked this test. Two services flunked this test: Orcad.js and – Tesseract. The purpose this first easy task was to wheat out any essentially not working online OCR services. And as the newspager name suggests the font is straight forward -Times New Roman. In the first round, the OCR input is a screenshot of the article – the image quality can not get better. The first four tests use the New York times article as input. You find the original input images by clicking on the preview image in the table headers - or see the reference section below. OCR Test 1: Recognize text in screenshot A New York Times article served as benchmark for OCR recognition quality in tests 1-4. The test results were not what you would expect. Six documents that are gradually more difficult to recognize serve as OCR benchmark: A screenshot, two scans, a mobile phone camera picture and, as highlight, text of an XKCD comic and readings from the image of a gas meter. This market overview is all about finding the best online service to convert images to plain text. Usability, speed, formatting, non-English language support are not rated. No-name OCR beats Google Docs OCR is just one of the surprising test results. What is the best free optical character recognition (OCR) service to convert (text in) images to plain, editable text? This review compares the recognition accuracy of free and commercial cloud OCR offerings. Update May 1, 2015: (a9t9) launched its very own free and open-source Online OCR service - try it out and let us know how it compares.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |