Deepstream facial landmarks with far faces

Accelerated Computing Intelligent Video Analytics DeepStream SDK
, , ,
1

Hello,

With the facial landmarks demo app I’m trying to retrieve the landmarks to post-process it but I noticed that if the detected face is too far (3 meters or more) the keypoints coordinates become too close from each other and i’m not able to process it.

Am I using the tool correctly ?

Hardware platform : Jetson AGX ORIN DevKit
Deepstream 7.0
Jeckpack 6
TensorRT 8.6.2.3
Cuda 12.2
NVIDIA GPU Driver version 540.3.0

3

Our demo just shows how to use the feature. If you have special scenario, like far faces, you need to retrain your own model to meet your special scenario.

4

This topic was automatically closed 14 days after the last reply. New replies are no longer allowed.