zihaomu

Hi @IamSVP94, can you attach exported ONNX file to the issue? It may be caused by the old Div operator not supporting broadcast. Recently, we have supported full broadcast in...

## Discussion of FP16's backend API There are two options: 1. FP16 is used as `DNN_BACKEND_OPENCV_FP16`. And this is what I have used now. And the following code is how...

Thank you for your answer @alalek, below is my reply. > What is about `net.setPreferableTarget(DNN_TARGET_CPU_FP16);`? `net.setPreferableTarget(DNN_TARGET_CPU_FP16);` is good choice, and the implementation of `DNN_TARGET_CPU_FP16` will not conflict with `DNN_BACKEND_CPU`. My...

Hi @alalek, thanks for mentioning issue #16633, it was very helpful. More precise description, what I want to do with this PR is just Default Mixed-precision:fp16 for convolutions and fp32...

Hi @asmorkalov, done.

Hi @the-star-sea. Since in OpenCV the `DB` has been supported by High-Level API, can you provide more speed and accuracy test data for DB and proposed ppocr?

@the-star-sea the ppocr's speed is super fast. two questions: 1. Are the pre-process and the post-process of ppocr's the same as the DB net? 2. Is there accuracy test result?

Thank you @the-star-sea! The accuracy result is good. Which validation data are used? How many data it has?

Thanks for the clarification. The `DB model` has hidded the post-process in the high level api, some thing like the following: ```py model = cv.dnn_TextDetectionModel_DB( cv.dnn.readNet(self._modelPath) ) # time start...

> 47628 That's a great answer. Thanks! Can we re-use the High-Level API of DB for PPocr with no change or little change? ```py model = cv.dnn_TextDetectionModel_DB( cv.dnn.readNet(modelPath_ppocr) ) model.detect(image)...