DIGIT with docker

I used this web(jetson-inference/digits-setup.md at master · dusty-nv/jetson-inference · GitHub) to install Nvidia Digit with docker.But I used this command line(nvidia-docker run --name digits -d -p 8888:5000 -v /home/bmw/data:/data:ro-v /home/bmw/digits-jobs:/workspace/jobs nvcr.io/nvidia/digits:18.05) on terminal is not working. How can I solve it?
And if want to train googlenet with pytorch on DIGIT, how can I do?

Hi,

Do you run the command on a desktop device?
Please noticed that this tutorial is for the host. It cannot be used on the Jetson Nano.

Thanks.

I have applyed the password and username on docker, but I can’t login. How can I solve it?

Yes, I used PC to login docker.

Hi @andy8902, when you docker login nvcr.io, you don’t input your normal NGC username/password. You will use $oauthtoken as the username and your API key as the password.

To generate your API key, see here:

If you want to train with PyTorch, see this part of the tutorial: GitHub - dusty-nv/jetson-inference: Hello AI World guide to deploying deep-learning inference networks and deep vision primitives with TensorRT and NVIDIA Jetson.

Note that the DIGITS part of the tutorial is now deprecated in lieu of using PyTorch directly.

I have logged in to docker for the first time. Every time I log in, do I have to enter the same name and password as the first time?

Hi,

Yes, please.
If the API key is somehow missed, you can also regenerate a new key via the link shared above.

Thanks.

When I use this web to login docker sucessfully, I want to use docker to enter DIGIT. When I inputed command line on top picture,it appeared information on buttom picture. How to solve it?

Are the command line shows that “/home/username/data” and “/home/username/digits-jobs” file paths?

Hi,

Please remember the space between ... 8888:5000 \ and -v ....

$ nvidia-docker run --name digits -d -p 8888:5000 \
 -v /home/username/data:/data:ro
 -v /home/username/digits-jobs:/workspace/jobs nvcr.io/nvidia/digits:18.05

Or you can concatenate them into one line command:

$ nvidia-docker run --name digits -d -p 8888:5000 -v /home/username/data:/data:ro  -v /home/username/digits-jobs:/workspace/jobs nvcr.io/nvidia/digits:18.05

Thanks.

I use this web(jetson-inference/imagenet-custom.md at master · dusty-nv/jetson-inference · GitHub)to download the network on Jetson nano. I encounter a problem below picture. Do I need to install Caffe on Jetson? or how to solve the problem? Thanks for your help.



Hi,

The command indicates to export the corresponding path to the $NET environment variable.
It should look like this:

$ export NET=/home/bmw/jetson-inference/Goo

You can check if the variable is correctly set via this command:

$ echo $NET

Thanks.

When I use this web(jetson-inference/imagenet-custom.md at master · dusty-nv/jetson-inference · GitHub) to detect object with camera(logitech C270) on jetson nano, it can’t attach camera below the picture. How can I solve the problem?

I use this command line :

./imagenet-camera.py /dev/video0
–prototxt=$NET/deploy.prototxt
–model=$NET/snapshot_iter_120.caffemodel
–labels=$NET/labels.txt
–input_blob=data
–output_blob=softmax

Hi,

You will need to check which node the camera is mounted first.
The simplest way is to run following command before/after the camera is attached.

$ ll /dev/video*

The please update the camera node of this command:

./imagenet-camera.py /dev/video0
–prototxt=$NET/deploy.prototxt
–model=$NET/snapshot_iter_120.caffemodel
–labels=$NET/labels.txt
–input_blob=data
–output_blob=softmax

Thanks.

I used your instruction to run with camera. It also can’t run.

Hi,

Based on the log, the video node is correct but the camera type is wrong.
The default camera is set to the nvarguscamerasrc which targets for the on-board camera.
For USB camera, please use v4l2src instead.

Not sure which jetson-inference source do you use.
If you clone it over 2 weeks ago, it’s recommended to update the source to the latest first.
The camera can automatically choose via /dev/video0 or csi://0 recently so you don’t need to manually update it.

Thanks.

https://elinux.org/Jetson_Nano#Cameras

But I used Logitech C270 on Jetson Nano. I want to ask a question what driver didn’t on Jetson nano?

Hi,

For kuanching USB cameras, please refer to