DNN - Text Detection and Recognition

Text detection and recognition nodes can detect where text is in an image and then provide the output with bounding box position and text string.

Pass the output of text detection into the input of text recognition to enable the full process.


See Deep Learning in OpenCV for information about supported frameworks and supported layers.

Some tested models with parameters:

Text Detection:

ModelSize (W x H)MeanScaleRGB
DB_IC15_resnet501280 x 736122.68, 116.67, 104.010.00392true
DB_IC15_resnet181280 x 736122.68, 116.67, 104.010.00392true
DB_TD500_resnet50736 x 736122.68, 116.67, 104.010.00392true
DB_TD500_resnet18736 x 736122.68, 116.67, 104.010.00392true

Text Recognition:

ModelVocabularySize (W x H)MeanScaleRGB
crnnalphabet 36100 x 32127.5, 127.5, 127.50.0078431false
crnn csalphabet 94100 x 32127.5, 127.5, 127.50.0078431true
crnn cs CNalphabet 3944 Chinese characters100 x 32127.5, 127.5, 127.50.0078431true