CUDA toolkit -Jetack 4.6.1

Robotics & Edge Computing Jetson & Embedded Systems Jetson Nano
1

Which version of CUDA toolkit is compatible with Jetack 4.6.1 on the jetson nano?

2

Hi @contactnikhilrb, CUDA 10.2 is compatible with JetPack 4.6.1, which is what it ships with in the SD card image.

4

@dusty_nv So are none of the later versions supported on JetPack 4.6.1?

5

No they are not, also Jetson Nano and JetPack 4 are EOL.

6

@dusty_nv Thank you for the revert.
I have one follow-up question: I’ve implemented a facial detection + recognition code on the Jetson Nano, the facial detection part is implemented using MediaPipe 0.8.5 (BlazeFace model) and runs on the GPU of the Jetson Nano, while the recognition is implemented using DLib’s face_recognition library and runs on the CPU of the Jetson Nano.

Commenting out the facial recognition part of the code, the facial detection part of the code runs at about 35 fps. But, when I try running both the detection+recognition the overall frame rate drops to about 5 fps.
Need help with how I can avoid this drastic drop in the overall frame rate on the Jetson Nano. Can I migrate Dlib’s facial_recognition library to the GPU (by compiling it with CUDA support) on JetPack 4.6.1, if yes, would this help? Also, please suggest alternative approaches to carry out facial recognition on the Nano that would run at a faster frame rate compared to Dlib’s facial_recognition library.

7

Sure @contactnikhilrb, I’m not personally familiar with DLib so you may want to create a new post about this, but after a quick search I found this and it appears that it can be compiled with CUDA enabled using the -DDLIB_USE_CUDA=1 flag (I would skip the conda part of this)

Also I remember this project for Jetson Nano that was doing face recognition:

8

@dusty_nv So would step 4 be sufficient to compile DLib with CUDA support from the link you shared?

$ git clone https://github.com/davisking/dlib.git
$ cd dlib
$ mkdir build
$ cd build
$ cmake .. -DDLIB_USE_CUDA=1 -DUSE_AVX_INSTRUCTIONS=1
$ cmake --build .
$ cd ..
$ python setup.py install --set DLIB_USE_CUDA=1
9

@dusty_nv Could you please confirm if dlib with CUDA is supported on JetPack 4.6.1?

10

@contactnikhilrb DLib is third-party software and I haven’t used it, so I’m not sure. If you encounter issues compiling it with CUDA enabled, please feel free to open a new forum topic and possibly also ask on their GitHub.

11

Linking to @contactnikhilrb’s other topic about DLib for others who may find this post:

13

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