Hi, I am using android_camera for detecting inference time. But I find that inference by android_camera app is much slower than using rpc remote call. How is that happen? Is that because at the same time, my phone is inference as well as dealing with some button and picture selection?
Android-camera app is directly downloaded and compiled from
And rpc call is used like this one https://tvm.apache.org/docs/tutorial/cross_compilation_and_rpc.html Both using same model and device.