Open vnnw opened 6 years ago
That would be a great toolkit. Thanks for the feature request! My only concern is that it would be hard for users to provide the necessary training data to train such a model. If you could tell us more about your use case and your data, that would be really useful for us.
If you don't need to prime the model for your own data, you can also try using the Vision framework, which has built-in OCR capabilities: https://developer.apple.com/documentation/vision/detecting_objects_in_still_images
@gustavla Thanks for the kind reply and suggestions. I am working for a e-commerce company. So my usage scenario mainly involves bank card recognition( card number, card holder' name etc), as well as illegal text description detection on product image (such as bait contact information to lure visitors to trade off the site) , the texts are multiple lingual in Chinese, English etc.
Will Turicreate implement algorithms like ctpn + crnn for scene text detection and recognition?