Hi all, we’re using the Arducam KingKong CM4 AI camera kit with a Raspberry Pi in an industrial setup. It was working fine until I ran ‘sudo apt update’ and ‘sudo apt upgrade’ on the Pi. After that, the camera isn’t detected anymore — the system just doesn’t see it. Anything I can try to bring it back?
This sounds like the known issue where running sudo apt update / sudo apt upgrade on the existing KingKong image breaks camera detection. The way to recover is to re-flash the SD card with the official Arducam KingKong image from https://sourceforge.net/projects/kingkong/ . One important note: after you flash this image, do NOT run sudo apt update or sudo apt upgrade on it.
Thanks for that. One thing I’m not sure about — the SD card sits inside the camera housing, not in an exposed slot. Do I need to open the enclosure to get the SD card out? I just don’t want to damage anything by opening it the wrong way.
Yes, you’ll need to open the enclosure to reach the SD card. You can use a screwdriver to unscrew the housing and take the SD card out of the Pi, then flash the KingKong image onto it. That’s the normal way to access it.
I’ve gone ahead and flashed the image, but the camera is still not working — still not detected. I’ve attached a screenshot of what I’m seeing. On a separate note: is the device able to capture data while it’s processing, for example send a command to capture the current image? Is there an API for that? Any update on this issue?
On the camera still not being recognized even with the new image — please make sure you’re flashing the KingKong image from https://sourceforge.net/projects/kingkong/ and that you haven’t run sudo apt update / sudo apt upgrade after flashing. Another customer reported success after re-flashing with that image, so I’d try that first. For the capture question, can you clarify what type of data you want to capture during processing?
Sorry for the confusion. The AI camera can open and run a pretrained model. What I want to know is: when the model detects something, is there output data available — like an object label such as ‘person’ — that I can grab from the backend and process however I need?
Got it. Yes, that detection output (object labels like ‘person’) can be captured and post-processed. Look at the rpicam-apps post-processing stages here: rpicam-apps/post_processing_stages at main · raspberrypi/rpicam-apps · GitHub , and there’s more detail in the official docs: https://www.raspberrypi.com/documentation/computers/camera_software.html#post-processing-with-rpicam-apps . To sum up: for the camera not being detected, stick with the KingKong image and avoid apt update/upgrade; for capturing detection results, the rpicam-apps post-processing stages are the recommended way to do it.