Recognizing text in images
The Vision framework has been improving its detection of text in images since its first iteration. In this section, we are going to learn some state-of-the-art techniques to obtain the best results on iOS 14.
We saw in the previous section that text detection in Vision can happen in two different ways, as defined by the value of recognitionLevel that we specify in the request: .fast and .accurate. Let's see the differences:
- Fast recognition: This uses character recognition to detect text character by character within a bounding box. It is optimized for real-time recognition and uses a smaller memory footprint than
.accurate. It doesn't handle rotated text or different fonts as well as the.accuratemethod. - Accurate recognition: This uses a neural network to detect strings and full lines, and then recognizes the words and sentences. By using a neural network and identifying words, the framework can detect or correct observations for...
