For excellent and fast face detection

I have tried the jetson-inference example.
I tried “Detectnet-camera facenet” but it did not come out as much as I wanted.
We considered DIGIST to study the corresponding facenet model.
In addition, I tried face recognition using haar or ResnetSSD through opencv.
As much as I wanted, but the frame rate dropped.
Since opencv’s ffmpeg uses cpu, the frame rate is expected to drop.
Can not change opencv’s input to gstreamer?
What are some ways to achieve good face detection?


You can try to find some state-of-the-art face detection model and run it with jetson_inference or TensorRT.

Our sample targets for demonstrating so the model may not be the best.
You can check the community to find other face models. For example:


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