I want to use C++ to invoke the trained model for detection, but there is an error in initialization during detection. Referring to the way the SSD calls the model in C++, the model requires a deploy conversion. So does the faster RCNN model need to be transformed as well? How do you convert?
Not all steps for inference are written with mxnet native operators. Gluon-CV model is actually written with only native operators so exporting symbol is possible.
I want to use C++ to invoke the trained model for detection, but there is an error in initialization during detection. Referring to the way the SSD calls the model in C++, the model requires a deploy conversion. So does the faster RCNN model need to be transformed as well? How do you convert?
Hope for help!