Vision Processing
#1
I'm not sure if this is better posted in hardware or software, but...

How does the Pine A compare with the Raspberry Pi in terms of vision processing?  I tried a simple OpenCV blob tracking example with the Pi, and it gave an unacceptable framerate of 5-7 FPS.  Can we expect a better framerate with the Pine (20-30 FPS)?
#2
The CPU in the Pine64 should be better than the RPi.
Wait a few weeks for benchmarks.

Get the A64+ that has a MIPI CSI camera interface. It would get much better performance than a USB camera.
#3
(12-15-2015, 03:37 AM)taros Wrote: The CPU in the Pine64 should be better than the RPi.
Wait a few weeks for benchmarks.

Get the A64+ that has a MIPI CSI camera interface. It would get much better performance than a USB camera.

Anybody actually receive one and try this with the 5mp camera. This is one of the things I wanted to do with this board. Smile
#4
(12-10-2015, 09:03 AM)codetheweb Wrote: I'm not sure if this is better posted in hardware or software, but...

How does the Pine A compare with the Raspberry Pi in terms of vision processing?  I tried a simple OpenCV blob tracking example with the Pi, and it gave an unacceptable framerate of 5-7 FPS.  Can we expect a better framerate with the Pine (20-30 FPS)?

Which pi?
#5
(03-01-2016, 09:28 PM)patrickhwood Wrote:
(12-10-2015, 09:03 AM)codetheweb Wrote: I'm not sure if this is better posted in hardware or software, but...

How does the Pine A compare with the Raspberry Pi in terms of vision processing?  I tried a simple OpenCV blob tracking example with the Pi, and it gave an unacceptable framerate of 5-7 FPS.  Can we expect a better framerate with the Pine (20-30 FPS)?

Which pi?

Good question. The pi 3 probably does laps around the older pi. 
It looks like there is an openCV release for android and for Linux. So, we should be able to run it in either case on the pine. http://opencv.org 
ibwonder if either package is for 64 bit. 
I also wonder whether we really gain a lot being 64 bit or not. I also wonder if having the camera with parallel communication on the pine is any better than serial on the pi. No idea. Anyway, anxious to try this. I'm surprised there's not more talking about this on the forum.
#6
Heck, the Pi 2 runs rings around the original Pi.

64-bit doesn't really buy you anything in terms of performance unless you're working with 64-bit long long values (e.g., extended fixed precision, software rendering). Depending on the architecture and instruction set, switching from 32 to 64-bit code can make your program and/or runtime memory larger, due to things like memory alignment requirements.

The Cortex-a53 architecture does have performance improvements over earlier Cortex CPUs, so a lot of 32-bit code will run faster at the same clock rate.
#7
(12-10-2015, 09:03 AM)codetheweb Wrote: I'm not sure if this is better posted in hardware or software, but...

How does the Pine A compare with the Raspberry Pi in terms of vision processing?  I tried a simple OpenCV blob tracking example with the Pi, and it gave an unacceptable framerate of 5-7 FPS.  Can we expect a better framerate with the Pine (20-30 FPS)?

Note, in reading the datasheet for the pine64 camera (https://drive.google.com/file/d/0B0cEs0l...yUWYw/view), I discovered that the camera can only do about 15 fps at full resolution.  If you don't need full resolution, then it get a higher frame rate (720p at 30fps and VGA at 60fps).  Note, this is just for the camera itself, and not any bottlenecks due to communication or processing.
#8
(03-02-2016, 06:50 AM)MichaelMeissner Wrote:
(12-10-2015, 09:03 AM)codetheweb Wrote: I'm not sure if this is better posted in hardware or software, but...

How does the Pine A compare with the Raspberry Pi in terms of vision processing?  I tried a simple OpenCV blob tracking example with the Pi, and it gave an unacceptable framerate of 5-7 FPS.  Can we expect a better framerate with the Pine (20-30 FPS)?

Note, in reading the datasheet for the pine64 camera (https://drive.google.com/file/d/0B0cEs0l...yUWYw/view), I discovered that the camera can only do about 15 fps at full resolution.  If you don't need full resolution, then it get a higher frame rate (720p at 30fps and VGA at 60fps).  Note, this is just for the camera itself, and not any bottlenecks due to communication or processing.

Ewww. Sad
Come on! And that's on a parallel line too. 
It's like putting a Fiero engine in a Ferrari (smh).
Not ideal. I'll have to send an email to these guys.

(03-02-2016, 06:50 AM)MichaelMeissner Wrote:
(12-10-2015, 09:03 AM)codetheweb Wrote: I'm not sure if this is better posted in hardware or software, but...

How does the Pine A compare with the Raspberry Pi in terms of vision processing?  I tried a simple OpenCV blob tracking example with the Pi, and it gave an unacceptable framerate of 5-7 FPS.  Can we expect a better framerate with the Pine (20-30 FPS)?

Note, in reading the datasheet for the pine64 camera (https://drive.google.com/file/d/0B0cEs0l...yUWYw/view), I discovered that the camera can only do about 15 fps at full resolution.  If you don't need full resolution, then it get a higher frame rate (720p at 30fps and VGA at 60fps).  Note, this is just for the camera itself, and not any bottlenecks due to communication or processing.

To be fair, this is what the rpi can do on the older hardware (not pi3):
  • 2592×1944 1-15fps, video or stills mode, Full sensor full FOV, default stills capture

  • 1920×1080 1-30fps, video mode, 1080p30 cropped

  • 1296×972 1-42fps, video mode, 4:3 aspect binned full FOV. Used for stills preview in raspistill.

  • 1296×730 1-49fps, video mode, 16:9 aspect , binned, full FOV (width), used for 720p

  • 640×480 42.1-60fps, video mode, up to VGAp60 binned

  • 640×480 60.1-90fps, video mode, up to VGAp90 binned
    https://www.raspberrypi.org/blog/new-cam...-released/ 
But the camera is more expensive (like 20 bucks).
#9
I plan to use mt Pine with OpenCV for some playing around with image processing as well.


Working with OpenCV, I found hugh differences in processing times depending how you write the code. I write in Python and was doing tracking stuff. Initially my code ran in 2 seconds per image, now have it down to 0.1 s, so there might be some room for optimization.
#10
(03-06-2016, 04:29 PM)NWUK Wrote: I plan to use mt Pine with OpenCV for some playing around with image processing as well.


Working with OpenCV, I found hugh differences in processing times depending how you write the code. I write in Python and was doing tracking stuff. Initially my code ran in 2 seconds per image, now have it down to 0.1 s, so there might be some room for optimization.
Hello I am planning to do some Image Processing on Pine through OpenCV. I need to ask some questions. Can you reply me on my Email p146385@nu.edu.pk. It would be a great help if you do reply. Thanks Smile


Forum Jump:


Users browsing this thread: 1 Guest(s)