Not able to run face recognition on GPU of Jetson nano

Robotics & Edge Computing Jetson Systems Jetson Nano
1

Installed dlib and face_recognition as per solution of Installing dlib with python3.6 in Jetson Nano .
I’m able to run face_recognition on cpu. but when I try to offload to GPU it gives error.
import face_recognition
image = face_recognition.load_image_file(“/home/batsy/Downloads/my_face.jpg”)
face_locations = face_recognition.face_locations(image)
the above code works. The following code fails:-
import face_recognition
image = face_recognition.load_image_file(“/home/batsy/Downloads/my_face.jpg”)
face_locations = face_recognition.face_locations(image, model=“cnn”)

Following is the error:-
File “trial_0.py”, line 3, in
face_locations = face_recognition.face_locations(image, model=“cnn”)
File “/home/batsy/.local/lib/python3.6/site-packages/face_recognition/api.py”, line 119, in face_locations
return [_trim_css_to_bounds(_rect_to_css(face.rect), img.shape) for face in _raw_face_locations(img, number_of_times_to_upsample, “cnn”)]
File “/home/batsy/.local/lib/python3.6/site-packages/face_recognition/api.py”, line 103, in _raw_face_locations
return cnn_face_detector(img, number_of_times_to_upsample)
RuntimeError: Error while calling cudnnFindConvolutionBackwardDataAlgorithm( context(), (const cudnnFilterDescriptor_t)filter_handle, descriptor(dest_desc), (const cudnnConvolutionDescriptor_t)conv_handle, descriptor(data), num_possible_algorithms, &num_algorithms, perf_results.data()) in file /tmp/pip-build-yrtsvnza/dlib/dlib/cuda/cudnn_dlibapi.cpp:856. code: 4, reason: A call to cuDNN failed
cudaStreamDestroy() failed. Reason: the launch timed out and was terminated
cudaFree() failed. Reason: the launch timed out and was terminated

2

Make sure to use the latest version 19.21 of Dlib, older versions are incompatible with JetPack 4.5.

4

Hi,

It seems that you are facing a similar issue as below topic:
https://github.com/ageitgey/face_recognition/issues/992#issuecomment-650533909

The user fixes the error by switching the library import order.
Could you also give it a try?

Thanks.

6

Hi @AastaLLL ,
Thanks for the reply. this solution worked.

7

Getting the below error while training the images ,

Traceback (most recent call last):
File “generate_training_data.py”, line 86, in
dets = face_detector(image, 1)
RuntimeError: Error while calling cudnnFindConvolutionBackwardDataAlgorithm( context(),

import order tried not working
import os
import numpy as np
import os
import dlib
import shutil
import argparse
import random
import sys
import face_recognition_models
import cv2

any idea why ?

using jetson-nano 4gb
dlib-19.22 (cuda enables)
opencv 4.5.1 (cuda enabled)

8

Hi gopurajesh,

Please help to open a new topic. Thanks