I hope someone can help me out. Following Tensorflow official documentation for object detection by means of transfer learning I have created customs model and I can detect my trained object in video, pictures, etc. via TF code. I have used one of pre-trained SSD mobilnet models and saved the final model as saved_model.pb . I would like to open the learned model in opencv DNN library for further processing - I am about to track the object with Kalman's filter.
How can I transfer saved_model.pb to any of readible formats of opencv dnn library?
I has gone through the many articles, official TF documentation, stackoverflow questions but did not find the proper way or I am doing something wrong ... I have tried to save as .h5 and than to pb, pbtxt, several environments with different versions of TF, opencv, etc. ... no success. Can anyone help me or guide me to some resource with information?