Thanks everyone for the feedback – much appreciated! ![🙏]()
Just to clarify why I’m emphasizing the use of hardware encoding (HW encoder) and aiming for low CPU usage, regardless of the camera model:
The goal is to build a final product that is efficient, compact, and user-friendly, especially for non-technical end users.
USB cameras are the preferred interface in my case, not CSI. That’s mainly because:
1- USB (2/3) is widely available and supports a large variety of cameras.
2- USB is much easier to connect/disconnect for end users compared to fragile ribbon cables (like CSI).
3- My casing and design are optimized for USB — it's plug-and-play.
Now, regarding keeping CPU usage low:
On CM4, using the hardware H.264 encoder lets me stream/record video with very low CPU usage, low system temperature, and reduced power consumption.
If I switch to a board like CM5 or Raspberry Pi 5, which lacks a dedicated GPU encoder, I’d have to rely on the CPU to encode video. Yes, it works — but it:
1- Uses significantly more CPU,
2- Raises the system temperature,
3- Increases current draw (not ideal for compact, embedded systems).
This matters a lot when designing for small enclosures with passive cooling and limited power budgets.
As for the camera module itself: I'm not strictly tied to any one model, but I’ve tested this particular USB camera and it works well on CM4 with the hardware encoder enabled. I don’t want to recompress streams in software (e.g., YUYV to H.264 via CPU), because that leads to quality loss and unnecessary resource use — as 6by9 mentioned.
Lastly, I understand CSI cameras like the IMX462 STARVIS work well with Picamera2 and offer great quality — and yes, they’re efficient. But again, this project needs USB camera support for practical and user experience reasons.
I'll attach some photos of the final product using CM4 and USB for context.
![Image]()
![Image]()
![Image]()
![Image]()
Thanks to bensimmo, fchk, and 6by9 for the excellent insights. — really appreciate the time and knowledge shared here. It’s always great to hear from people with hands-on experience, especially with different hardware setups and real-world applications.
regards
Just to clarify why I’m emphasizing the use of hardware encoding (HW encoder) and aiming for low CPU usage, regardless of the camera model:
The goal is to build a final product that is efficient, compact, and user-friendly, especially for non-technical end users.
USB cameras are the preferred interface in my case, not CSI. That’s mainly because:
1- USB (2/3) is widely available and supports a large variety of cameras.
2- USB is much easier to connect/disconnect for end users compared to fragile ribbon cables (like CSI).
3- My casing and design are optimized for USB — it's plug-and-play.
Now, regarding keeping CPU usage low:
On CM4, using the hardware H.264 encoder lets me stream/record video with very low CPU usage, low system temperature, and reduced power consumption.
If I switch to a board like CM5 or Raspberry Pi 5, which lacks a dedicated GPU encoder, I’d have to rely on the CPU to encode video. Yes, it works — but it:
1- Uses significantly more CPU,
2- Raises the system temperature,
3- Increases current draw (not ideal for compact, embedded systems).
This matters a lot when designing for small enclosures with passive cooling and limited power budgets.
As for the camera module itself: I'm not strictly tied to any one model, but I’ve tested this particular USB camera and it works well on CM4 with the hardware encoder enabled. I don’t want to recompress streams in software (e.g., YUYV to H.264 via CPU), because that leads to quality loss and unnecessary resource use — as 6by9 mentioned.
Lastly, I understand CSI cameras like the IMX462 STARVIS work well with Picamera2 and offer great quality — and yes, they’re efficient. But again, this project needs USB camera support for practical and user experience reasons.
I'll attach some photos of the final product using CM4 and USB for context.
Thanks to bensimmo, fchk, and 6by9 for the excellent insights. — really appreciate the time and knowledge shared here. It’s always great to hear from people with hands-on experience, especially with different hardware setups and real-world applications.
regards
Statistics: Posted by mot1639 — Mon Jun 02, 2025 3:00 am