Google Cloud Vision AI

Vision AI is Google Cloud's set of image and video understanding services, reached through APIs rather than installed software. Pretrained models extract information from images, documents and video, and the same platform supports training a custom model on your own labeled images, with no code training in a managed environment, when the pretrained set does not cover a task. The Vision API covers label detection, optical character recognition for both general text and dense documents, face detection, landmark and logo recognition, object localization, image properties, crop hints, safe search moderation and web detection that matches an image against pages online. Typical uses are classifying and tagging large image libraries, pulling structured fields out of scanned forms and receipts, moderating uploaded content and adding visual search to an application. Each page of a multipage file such as a PDF counts as a separate image. Billing is per unit, where one feature applied to one image is one unit. The first 1,000 units per feature per month are free. From 1,001 to 5 million units a month, most features cost $1.50 per 1,000 units, object localization costs $2.25 and web detection costs $3.50. Above 5 million units a month most features drop to $0.60 per 1,000 units, label detection to $1.00 and object localization to $1.50, while web detection moves to a rate arranged with Google. Charges are calculated in blocks of 1,000 units with the final partial block prorated, and new customers receive up to $300 in free credits. Any other Google Cloud resources used alongside it, such as storage and compute, are billed separately.