example face_detection_and_landmarks.sh failed

General
1

Hello,
I am using hailo 8 m.2 Key E module on an Intel N305 platform.
I installed tappas 3.31.0 docker and I tried to run the gstreamer examples:

  • The example object_detection_and_pose_estimation.sh run OK
  • The example face_detection_and_landmarks.sh runs but it does not displays face detection/landmarks, it only displays original input video of a classroom with people.

Am I missing something?

2

Hi, welcome to the community.
This is due to a licensing issue we found out lately regarding the facial landmarks network.
We had to remove some required npy parameters from our servers.
We are working to replace this on another open source network.
What is your use case?
If you need more sparse facial landmarks (5 points) you can use the “scrfd_10g” network which is running face detection with simple landmarks, see example here: tappas/apps/h8/gstreamer/general/face_recognition at master · hailo-ai/tappas · GitHub

Another good resource is this project: Accelerating the MediaPipe models on Raspberry Pi 5 AI Kit - Hackster.io
It got hand and facial landmarks.

3

If you are using the mediapipe project note that there are some issues you need to be aware of.
First build virtualenv using site packages.

python -m venv venv_blaze --system-site-packages
source venv_blaze/bin/activate

Install dependencies:

pip3 install tflite_runtime matplotlib plotly kaleido numpy==1.24

Download the models according to the description.

To run the facial landmarks use this:

python3 blaze_detect_live.py -b face