Media Server can detect regions, in images and video, that contain text.

You can also use text detection to identify vehicles (by finding the text on the vehicle's number plate), and then identify the vehicle's make, model, and color (see Vehicle Make and Model Recognition). You should only use text detection for this purpose if you cannot use number plate recognition, for example when the number plates exhibit out of plane rotation, because text detection identifies all text in a frame rather than just the number plates.

Text detection does not read the text. If you want to detect and recognize text, you can use OCR, or run text detection to locate regions that contain text and then run OCR on those regions. In cases where the media contains small amounts of text within a scene (and you would use OCRMode=scene), using text detection to find the text and then running OCR can provide better performance than running OCR on the entire image or video frame. Where there is a large amount of text, for example an image of a document, using OCR (with OCRMode=Document) is the best approach.

In some cases text detection might be able to detect text that cannot be read by OCR or number plate recognition, for example due to out-of-plane skew. You can use text detection to ensure that the text is captured, to be read and evaluated by a person. This might be useful for saving images or video for evidential purposes, for example.