Jetbot camera/jupyter slows down

Hi everyone,

I have been going through the tutorials for jetbot. I have noticed that as I have worked through the data_collection notebook for object detection that the camera image intailizes fine initially, but then gradually begins to lag and then becomes slower and slower until the lag becomes a minute or more behind.

The notebook in general becomes slower and take several minutes for one cell in the tutorial to execute.

I thought that this was due to power consumption, but I have plugged into the usb wall adapter and the same thing occurs.

If I reset the entire JETSON Nano not just jupyter it starts working fine again, but then gradually slows down.

No error messages are thrown. Just everything slows down.

Hi pettinger.michael,

Thanks for reaching out!

I’m not immediately sure what the cause of this might be. Do you mind sharing which JetBot hardware platform, a s well as which JetBot SD card image you are using?

Best,
John

Hi John @jaybdub I have JETSON nano 2GB. The image on the SD card is from the zip file
jetbot-042_nano-2gb-jp441.zip downloaded 12/22/2020.

The Jetbot was assembled from the www.jetbot.org instructions.

I’m using IMX219-160 camera.

Would you pls check Jetson Nano’s power mode?
$ sudo nvpmodel -q

@MtHiker Output is:
NVPM WARN: fan mode is not set!
NV Power Mode: MAXN
0

I believe this is set correctly. I do not have a fan attached.

@jaybdub I was wondering if there was any further development on this. I haven’t been able to move forward without getting a fix to this.

Similar issue that I’m facing. but mine even worse. I never get notebooks running fast from get-go. Every time when I started the Jupyter notebook from web page, it takes at least 5 mins to refresh the workspace and show up the main page.

The image display is lagging by 30secs to a few minutes.

I set the power mode to MAX (10W), using wall ps. I ping Nano IP address from host PC, the average traffic round trip time is <10ms.

I tried both jetbot images with jetpack v4.4 and v4.5, but same results.

Is there anything to improve? This extremely slow operation makes experimental codes impossible to run meaningfully.
Thanks!
Terry

Down the camera fps to 4:

before:
camera = Camera.instance(width=224, height=224)
after:
camera = Camera.instance(width=224, height=224, fps=4)

This performance is affected by network speed. (Jetson Nano <-> PC)
When the camera frame variable data is updated, traitlets displays it in the web browser.
Camera frame loading runs in a thread. When the display in the web browser is not complete, next task (frame to browser) will be buffered.
If the network is slow, the display in the web browser will be slow and the buffer will increase. This is why it appears slow.

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