MindOCR is an open source toolbox for OCR development and application based on the MindSpore framework, which can help users train and apply the industry’s best text detection and text recognition models, such as DBNet/DBNet++ and CRNN/SVTR, to meet the needs of image text understanding . Main features: Modular design: MindOCR decouples OCR tasks into multiple configurable modules. Users only need to modify a few lines of code to easily configure the whole process of training and evaluation on customized data and models; high performance : The pre-training weights and training methods provided by MindOCR can make it achieve competitive performance on OCR tasks… |
#OCR #development #application #toolbox #MindOCR