Hi everyone,
Suppose I want to use python inference for a deep model and do inference in plugin, one solution is to copy all regular inference python codes in do_transform_ip method in the class of gstreamer plugin, and convert buffer to numpy array and then pass to the model and get inference and create metadata for that buffer and then pass metadata to next element, but I don't know is it optimal solution convert buffer to numpy array for feeding the model? I want to know, in the c++ gstinfer element, the buffer converted to numpy array before feed to model? Is it possible to feed buffer to model? I want to know converting buffer to numpy array is has overload? I know converting buffer to numpy array in prob_function has overload, but I don't know has overload in the inside plugin overload also?
Hi everyone, Suppose I want to use python inference for a deep model and do inference in plugin, one solution is to copy all regular inference python codes in do_transform_ip method in the class of gstreamer plugin, and convert buffer to numpy array and then pass to the model and get inference and create metadata for that buffer and then pass metadata to next element, but I don't know is it optimal solution convert buffer to numpy array for feeding the model? I want to know, in the c++ gstinfer element, the buffer converted to numpy array before feed to model? Is it possible to feed buffer to model? I want to know converting buffer to numpy array is has overload? I know converting buffer to numpy array in prob_function has overload, but I don't know has overload in the inside plugin overload also?