Error with Python face_recognition

Install face_recognition:
sudo pip3 install face_recognition

Download demo:
git clone --single-branch https://github.com/ageitgey/face_recognition.git

run demo:

cd examples
python3 facerec_from_webcam_faster.py

When a face is detected, the array has an item and the error occurs.

/usr/local/lib/python3.6/dist-packages/face_recognition/api.py:222: RuntimeWarning: invalid value encountered in less_equal
return list(face_distance[known_face_encodings, face_encoding_to_check) <= tolerance)

Hi,

There is an cuDNN error on the Jetson Nano which leads to some issue in the dlib.
You can find more information in this topic:
https://devtalk.nvidia.com/default/topic/1049660/issues-with-dlib-library/

In summary, you will need to build dlib from source with the steps shared in this comment:
https://devtalk.nvidia.com/default/topic/1049660/jetson-nano/issues-with-dlib-library/post/5336330/#5336330

Thanks.

We have confirmed that this issue is fixed in the cuDNN v7.6.1.

how can I update to cuDNN v7.6.1.?

Hi,
I had the same error on the Jetson Nano with face recognition:
/usr/local/lib/python3.6/dist-packages/face_recognition/api.py:222: RuntimeWarning: invalid value encountered in less_equal
return list(face_distance[known_face_encodings, face_encoding_to_check) <= tolerance)

And I go to the download page, there isn’t cuDNN v7.6.1
https://developer.nvidia.com/rdp/cudnn-download

Please help, thanks.

Hi,

cuDNN 7.6 is not available yet.
Please wait for our future JetPack release to get the fix.

Currently, you can apply the workaround mentioned here first:
https://devtalk.nvidia.com/default/topic/1049660/jetson-nano/issues-with-dlib-library/post/5336330/#5336330

Thanks.

Hi,

I had the same error on the Jetson Nano with face recognition.
But, I have confirmed that face recognition works well on Windows and Raspberry Pi 3 B (with Python3).

Hi, i am also facing the same issue. its working in windows but not working in jetson nano

Hi nanos,
Works everywhere else but not on jetson nano yet.
Any progress in fixing this?
Thx+regards