Some problems with dlib-gpu and jetpack

[f25513115]

Could anyone tell me if dlib-gpu can work on jetson nano？
Firstly, I wanted to download CUDA10.1. But it was not compatible with ARM.
Then, I wanted to installed jetpack. Unfortunately, I still got AMD version. BTW, I read someone installed AMD version on nano successfully. So I confused about that.
I don’t have any idea about jetpack and even don’t know dlib-gpu can work or not. Becasue I read a post said that there were some problem with nano’s gpu. But that post was opend in 2020. Therefore, I haven’t known this problem was fixed yet.

[kayccc]

The dlib-gpu migth be not supported, need to confirm.

[f25513115]

Does that mean we just can use dlib-cpu on Nano?
Could you tell me how to install jetpack successfully?

[AastaLLL]

Hi,

We have confirmed that dlib-19.21 is working on the Jetson.
For other versions, you should be able to compile it in a similar way.

Simple accelerated face recognition

Hi, The fix is adding the cuDNNv8 new API support. And it is available in the newest dlib-19.21 release in Aug 8. So you can build it from source directly: Install dependencies $ sudo apt-get install python3-pip $ sudo apt-get install libjpeg-dev Build dlib from source $ wget http://dlib.net/files/dlib-19.21.tar.bz2 $ tar jxvf dlib-19.21.tar.bz2 $ cd dlib-19.21/ $ mkdir build $ cd build/ $ cmake .. $ cmake --build . $ cd ../ $ sudo python3 setup.py install $ sudo pip3 install face_recogni…

Thanks.

[f25513115]

Thank you very much~
So I should install jetpack first then following these steps, right?
Or when I finalize these steps, I’ll install cuda and cudnn in the same time?

[AastaLLL]

Hi,

The steps only compile and install dlib library.
Please install JetPack components first.

Thanks.

[f25513115]

Yes, thank you.
But I can install JetPack very well these days
So I would like to ask you if ypu have any recommand websites or steps can share me?

[AastaLLL]

Hi,

After you install all the JetPack components, please build dlib with the following instructions:

Simple accelerated face recognition

Hi, The fix is adding the cuDNNv8 new API support. And it is available in the newest dlib-19.21 release in Aug 8. So you can build it from source directly: Install dependencies $ sudo apt-get install python3-pip $ sudo apt-get install libjpeg-dev Build dlib from source $ wget http://dlib.net/files/dlib-19.21.tar.bz2 $ tar jxvf dlib-19.21.tar.bz2 $ cd dlib-19.21/ $ mkdir build $ cd build/ $ cmake .. $ cmake --build . $ cd ../ $ sudo python3 setup.py install $ sudo pip3 install face_recogni…

Thanks.

[f25513115]

sorry my mistake. I meant the steps about jetpack.

[AastaLLL]

Hi,

Below is the JetPack document.
For Jetson Nano, the latest software version should be 4.6.2:
https://docs.nvidia.com/jetson/jetpack/install-jetpack/index.html

Thanks.

[f25513115]

I found this page about jetpack4.2
https://developer.nvidia.com/embedded/jetpack-4_2
I think that’s a easy way to install jetpack(download sd card image with jetpack4.2) but I’m not sure the version is compatible with dlib19.21
I also see 4.6version can be installed by this way. I’ll try it

[kayccc]

Is this still an issue to support? Any result can be shared? Thanks

[f25513115]

The problem was fixed! Thank you very much.
I downloaded jetpack 4.6 version and installed dlib-gpu 19.21 successfully.

[kayccc]

Glad to know issue fixed, thanks for the confirmation.