Face detection on Jetson Nano

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

Hi

I tried to run the opencv face detection and recognition using dnn model res10_300x300_ssd_iter_140000.caffemodel but overall fps which I am getting is ~3 frames per second which feels a bit low.
I manually build and compiled the opencv with cuda support so I believe it should be faster but results are not good.
It looks like the nano does not support the opencv cv2.dnn functionality which makes the overall performance very low. Is there any alternative for that.?

Is there any other way through which we can achieve the face detection and recognition on jetson nano. Which model is recommended to run on nano for face detection.

Thanks

2

Hi,

Do you have the Caffe model?
If yes, it’s recommended to use TensorRT instead.

Here is a example for OpenCV → TensorRT:
[url]https://devtalk.nvidia.com/default/topic/1049958/jetson-nano/densenet121-transplanting-using-tensorrt/post/5336325/#5336325[/url]

Thanks.

3

Check out my TensorRT implementation of MTCNN face detector. It runs at roughly 4~8 FPS on Jetson Nano.

[url]https://devtalk.nvidia.com/default/topic/1050494/jetson-nano/opencv-face-detection-poor-performance-with-jetson-nano/post/5385407/#5385407[/url]