Use gmsl camera with python


As we want to use a ready trained tensorflow dnn we like to simple feed the images from the camera?
What is the best way or typical way to get this done?

My first intention was a simple python script feeding the gmsl camera stream but i am not able to read the images out. I tried to get a gstream pipline working but no luck so far.

I was inspired by this post:

gstream problem for now is this:
No element: nvcamerasrc

got a little confused by this post:

We use the drive px 2 with the latest firmware and ubuntu 16.04.

Some help would be appreciated thanks!

Hello tinokl,

JetsonTX1 and DrivePX2 are different.
Could you please refer to below link? Thanks.

Hello Steve,

I searched through the docs but I could only find some cpp related examples.
For that reason i searched the web and got some hints from jetson related projects.
(I know they are different but some tips worked for me)

For quick science driven development python has some benefits as i like to use tensorflow and keras.

By now i am searching for a simple solution to get the camera stream to my python script but
for future purposes i need them within ROS. Actually i would prefer a ROS ready solution, so we can grab the images wherever we need them.

In this post:
a guideline doc was mention, any new information on that?

ROS is working pretty out of the box except for the cameras.


Hello tinokl,

A upcoming PDK will be released in May. Thanks.

Hi tinokl,

There’s an update of ROS in

just wondering have you figured out how to get gmsl camera stream in python? If you did, can you please share some instructions? Thanks!

Hi WZhouuuu,

thanks for the update on ROS I already joined the conversation.

Unfortunaly i did not found a way to get the cameras working in python but i am still very interessted in a solution (thats why i am asking for a gstreamer pipeline or something similar).

If anybody has a working solution please give us some information.

Thanks in advance!