Safe Picamera2 upgrade for Pivistation 5 AI with 64MP Hawkeye camera

Raspberry Pi Cameras
1

Hi everyone, I’m running a Pivistation 5 AI with the integrated 64MP Hawkeye camera module on a Raspberry Pi 5, and I’m trying to upgrade Picamera2 to get access to the devices module (need 0.3.20+ for Hailo-8L inference integration). My current setup has picamera2 0.3.19 and HailoRT 4.17.0. I’m worried about breaking things since the system came pre-configured from Arducam. Can I safely upgrade picamera2 without messing up the 64MP Hawkeye support? I’ve been looking at the Picamera2 User Guide but I’m not sure what’s safe to touch.

2

Thanks for reaching out. Before I recommend anything, can I ask a few clarifying questions?

  1. What specific features or bug fixes are you hoping to get from the newer picamera2 version?
  2. Have you already tried any upgrade steps (apt upgrade, pip install --upgrade, etc.)?
  3. Is the Arducam stream service (the web-based preview tool) something you rely on?

I ask because the 64MP Hawkeye camera module on the Pivistation 5 AI requires a specifically tuned and patched version of the libcamera stack to support the ultra-high-resolution sensor. The pre-installed image ships with custom Arducam patches that are critical for the Hawkeye to function properly. Running standard upgrades could overwrite those patches and break camera functionality entirely.

3

Good questions. To clarify:

  1. I specifically need picamera2 ≥ 0.3.20 because earlier versions don’t include the devices module. That module provides an API for integrating Hailo inference directly within picamera2, which is what I’m trying to set up with the Hailo-8L. Currently on 0.3.19, importing from picamera2.devices import Hailo fails with an ImportError.

  2. I haven’t tried upgrading yet — that’s why I’m asking first. I read the warning in the Arducam docs about not breaking compatibility, so I wanted to check before doing anything.

  3. The Arducam stream service is nice to have but not critical. My main priority is getting the Hailo integration to work cleanly with picamera2 while keeping the 64MP Hawkeye functional.

What’s the recommended approach here?

4

Thanks for the details — that makes the situation clearer. Let me explain what’s going on under the hood:

The root cause here is that the Pivistation 5 AI image ships with a custom Arducam-patched libcamera stack. The 64MP Hawkeye sensor is an ultra-high-resolution camera that requires specific sensor drivers and patches that aren’t in the standard Raspberry Pi OS repositories. If you run:

  • apt upgrade (for picamera2/libcamera packages)
  • pip install --upgrade picamera2

…those standard package managers will overwrite the Arducam-specific patches, and the camera will stop working — or at best, be limited to lower resolutions.

That said, I don’t recommend upgrading picamera2 independently on the stock Arducam image. I can check with our R&D team to see if there’s a verified updated image or a safe manual upgrade path, but that would take some time.

Alternatively, there’s another approach you could take that avoids this conflict entirely.

5

I see, so the patches are the bottleneck. If there’s a safe upgrade path from R&D that would be ideal, but I’m also open to hearing the alternative. What are you thinking? I just need picamera2 >= 0.3.20 with the devices module + HailoRT compatibility while keeping the 64MP Hawkeye working.

6

Here’s the alternative solution I’d recommend:

Since you mentioned the Arducam stream service isn’t critical, you could move your development environment to Raspberry Pi OS Trixie (Testing). Here’s why:

Raspberry Pi OS Trixie ships with much more modern system libraries, including newer versions of libcamera and picamera2 that natively support the devices module. This avoids the entire conflict of overwriting custom patches because you’d be starting fresh on a clean OS, and the standard packages on Trixie are already at versions that support everything you need.

Here are the steps:

Step 1: Download the latest official Raspberry Pi OS (Trixie/Testing) from the Raspberry Pi website and flash it to a new SD card.

Step 2: Configure the 64MP Hawkeye camera using the updated software configuration guide:
https://docs.arducam.com/Raspberry-Pi-Camera/Native-camera/64MP-Hawkeye/#software-configuration

Step 3: On Trixie, the system libcamera and picamera2 packages are naturally at version levels that include the devices module. You won’t need to manually patch or upgrade anything — from picamera2.devices import Hailo should work out of the box.

Key insight: Moving to Trixie avoids overwriting custom patches on the older Arducam image while giving you the architectural benefits of the newer libcamera and picamera2 releases needed for Hailo-8L integration with the devices module.

If you absolutely need to stay on the Arducam image, let me know and I can check with R&D about a custom updated image — but Trixie is likely the faster and cleaner path forward.