Deepstream support for Face detction

• Hardware Platform (Jetson Xavier)
• DeepStream Version [5.1]
• JetPack Version (32.5.1)
• TensorRT Version [7.1]
• Issue Type( questions)

Hai, We are working on AI face detection and confirmation algorithem using Deepstream-SDK on Jetson Xavier.

We are new in TensorRT and Deepstream. We have installed Deepstream-5.1 and we can run the

While following the FaceDetectIR procedure and run the code, we got the below error.

    $ docker run --gpus all -it --rm -v /tmp/.X11-unix:/tmp/.X11-unix -e DISPLAY=$DISPLAY -v /home/nvidia/nfs/:/opt/nvidia/deepstream/deepstream-5.1/samples/models/tlt_pretrained_models \
    -w /opt/nvidia/deepstream/deepstream-5.1/samples/configs/tlt_pretrained_models \
    deepstream-app -c deepstream_app_source1_labels_faceirnet.txt

standard_init_linux.go:211: exec user process caused "exec format error"

In jetson-inference/data/networks/facenet-120/, we have caffemodel,labels and prototxt files.

we have configured the above files in dstest1_pgie_config.txt and run the test but not working.

Is any default deepstream application for face detection is available? or Please suggeust a steps to be follow for face detection from live streaming video using Deepstream?

We really appriciate your suggeustions.


“exec format error” usually indicates the container is compiled for other platform.

Please noted that is a desktop container.
For Jetson, please use instead.


Thanks for the reply AastaLLL. We followed Output Tensor in FaceDetectIR - #7 by Andrey1984 and used the app, now face detection in live camera is working fine.

Now we want to test face recognition. We have a set of pre-trained data set in .yml format. Please advise us how to use the custom data set in deepstream and recognise the face? or Is there any default application for face recognition in deepstream?

Thanks in advance.

@karthickdivw2 for face recognition you may like to try this Intel RealSense ID – Facial Authentication


Do you want to apply inference on a customized database?
If yes, you can check following sample for image source inferencing.