The Google Lens team built the OCR engine, which uses Google Search and the Knowledge Graph to train machine learning algorithms by scanning different characters, languages, fonts, and drawings from Google Books.
Google Lens, the artificial intelligence (AI) image-recognition technology Google unveiled last May, is now capable of identifying more than 1 billion objects after a year and a half of progress, the company said On Wednesday.
Google Lens is an image and text recognition technology based on AI and computer vision that allows users to take a photo of a recipe on their phone and add ingredients to their shopping list, or take a business card and add contacts to their list. To train Google Lens, the team built their own optical character recognition (OCR) engine, combined with language Knowledge accumulated through Google searches and the Knowledge Graph, The machine learning algorithms were then trained by scanning different characters, languages, fonts, and drawings from Google Books.
When it’s hard to decipher a 0 or an O, Lens uses Google’s search language and spelling correction model. Just as Google can change bannana to banana, Lens understands that c00kie probably means cookie.
Google says it now uses its OCR engine for other purposes, such as reading product labels, and Lens can now identify more than 1 billion products, four times as many as it did a year and a half ago.
In addition to reading text from images, Lens also adds the ability to quickly search and integrate image searches in the U.S. By pointing your phone at clothes, lights or other home decor items. Lens also displays similar items and offers price comparisons and comparisons.