Im on a rpi 3b doing some test on face tracking, im using face-detection-adas-0001 model and python.
I have a test using openCv and another one using inference_engine, and on the openCv version get almost double fps(7.7 fps vs 4.5fps) is that difference in performance correct? Is there any way of optimizing to get more fps?
The overall frame rate for Inference engine is rather determined only by the slowest part of the pipeline (decoding IR inference) and not by the sum of the stages which includes latency in capturing input frames.
However you may try improving the performance by using Inference Engine Async API.
Yes i tried and performance increased, but also make opencv async.... so ended in the same situation, opencv seems faster.
That seems extrange to me as i expected inference engine to be faster, thats why im asking
Can you please share with us the codes, commands and necessary files that you are using to load the model and inferencing using OpenCV and OpenVINO.
We don't have any public benchmarks between OpenVINO and OpenCV. However you may try optimizing the OpenVINO Inference engine using techniques given at smart video workshop.