I made a object detection android application but the frame rate is not going above 3-4. How can I improve the fps?
By reducing the image size to a multiple of 32, you can get a higher frame rate. Here comes the trade-off between Speed and Accuracy. You can reduce the image size until you receive satisfactory accuracy for your use-case. Thank you.
I just want to say that I am Using GPU, I attained 70 FPS with YoloV4-Tiny even with better accuracy that can be improved further. And, With CPU, I reached around 7 Frame Per Second Speed.
that depends on many aspects!
are you using the phones GPU?
are the input images to large?
is your model running completely on the GPU? (that depends on the operations it uses)
one way to know that is using the benchmark tool: https://www.tensorflow.org/lite/performance/measurement
hope this helps
Thanks for Information.