User avatar
jbeale
Posts: 3499
Joined: Tue Nov 22, 2011 11:51 pm
Contact: Website

Re: PiKrellCam: motion vector detect + OSD web interface

Fri May 06, 2016 12:21 am

Do we know if NFS on the Pi has the performance required to handle live video recording? Has anyone shown that to work?

Regarding accurate timestamps, I have noticed that the "burned-in" date/time on the video occasionally has a hiccup when it doesn't update for a second or more. Also, on my cameras the nominally 24 fps is more accurately 23.04 fps. I measured this using an external strobe light and adjusting the rate for minimum drift.

User avatar
Gavinmc42
Posts: 3916
Joined: Wed Aug 28, 2013 3:31 am

Re: PiKrellCam: motion vector detect + OSD web interface

Fri May 06, 2016 2:39 am

Pi B+, 2 or 3?
Do we know if NFS on the Pi has the performance required to handle live video recording? Has anyone shown that to work?
Seem to remember video over network is only 4-6MBS, at least for h.264 encoded playing.
h.264 probably the best as that is native to the Pi.
Pikrellcam has one of the fastest video over network already, much less lag than others I have tried.

Or do you want a Pi to be a NAS recording live video over a network?
Could one Pi 2 handle 4 x Pi B+?
MotionEye OS has that as an option, recording from network cameras or Pi setup as network camera.
Have not tested it, still trying to figure out best and/or easiest NAS software for a Pi

.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
Gavinmc42
Posts: 3916
Joined: Wed Aug 28, 2013 3:31 am

Re: PiKrellCam: motion vector detect + OSD web interface

Fri May 06, 2016 1:18 pm

Hi Billw,
It reads the FIFO every 1/10 sec. For defaults of 24fps and mjpeg_divider 4, it samples for motion detect every 1/6 sec
Yep found that and circular buffer of 35secs?
Might have to time walkers across frame, they keep disappearing.
It's only one compare to a minimum threshold. To get the min/max compare you want I really think jbeale is on the right track using the csv data as you want a decision based on final outcome of the video and a post process with the on_motion_end command makes sense for that.
Well you do provide the src code:)
First I checked it could compile by running make, yippee nothing broken.
Then I started to break it.
Added a magnitude_max variable and a compare less than test in motion.c after first pass.
Added a max limit etc to other files and it compiled fine and is running.
Now just have to go to bed and see what daylight brings, working or horribly stuffed.
I have set max vector limit to 150, hopefully it just works like normal until I start bringing max limit down.

The whole point is to reject large magnitude vectors videos before they get saved to SD card.
Sure I can run tests on the csv and probably should have done that by saving videos to ram disk then testing them.
Wish I had thought of that before hacking your code.
Still have to modify the php settings stuff.

Not sure if Jessie Lite has the compile tools, but it sure is nice hacking code, compiling and running remotely. One day all devices will be like this?

Love this software.
Tomorrow maybe not, fingers cross.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

jefferyanderson
Posts: 4
Joined: Thu May 05, 2016 12:01 pm

Re: PiKrellCam: motion vector detect + OSD web interface

Fri May 06, 2016 1:37 pm

billw wrote: The annotation string overlay is updated once per second, but the code could be modified to
update it more frequently and a subsecond option added. But how would you use that since it's
just a visual overlay?
Videos are started on iframes which are requested each time a system time second increments and
the system time is sampled 10 times per second. So I don't know if 100 msec accuracy could be
relied on, but videos should start close to 1 second boundaries. If the start times were logged
somewhere, could that plus a frame rate calculation get close to what you want?
Bill,
Thanks for the prompt reply! The visual overlay isn't a great solution, but since i"m doing OCR already on the image this could potentially be used. That said, it's a pretty painful/terrible way to solve it. The log file would be far better - would it somehow tie the filename to the start time?

Just so I can understand the operation better, does the fact the videos are to start +/-100ms on the one-second boundary have an impact on the video contents, i.e. could there be up to 900ms of delay if motion was detected but it's not yet on the 1 second boundary, or, alternatively, does this mean that the recorded time could be wrong by 900ms?

Thanks again for the great piece of software.
Jeff

billw
Posts: 403
Joined: Tue Sep 18, 2012 8:23 pm

Re: PiKrellCam: motion vector detect + OSD web interface

Fri May 06, 2016 6:37 pm

jefferyanderson wrote: Just so I can understand the operation better, does the fact the videos are to start +/-100ms on the one-second boundary have an impact on the video contents, i.e. could there be up to 900ms of delay if motion was detected but it's not yet on the 1 second boundary, or, alternatively, does this mean that the recorded time could be wrong by 900ms?

Thanks again for the great piece of software.
Jeff
Videos have to start on a video iframe and the way pikrellcam works now:

While waiting for a motion event:
Sample system time 10 times/sec. When system time second increment is detected, request a video iframe. When iframe video data is delivered, save a pointer to that video data in the circular buffer. Enough of these pointers are saved to cover the configured pre capture time and there should be one pointer per second.

When a motion event is detected:
Create the video file with the current time embedded in its filename. Get iframe pointer for current time minus pre capture time and write video data from that pointer to the current position in the circular buffer. Then carry on writing video data as it comes subject to post capture and event gap times. The actual pre capture time in the video will be the configured time plus the fraction of second between the last iframe received and the time the motion event occurred. At event gap expire, close the h264 file and use MP4Box to mp4 convert it.

What we have then is a video which has a filename embedding the system time of the first motion detect event with one second resolution. And the mtime of the video is when MP4Box finishes converting to mp4 - not useful for what you want


Since you brought this up, I'm now thinking of changing this:
I can record in the iframe table the GPU video frame time stamps as the iframes are seen. Then use the iframe time of the actual video start to create the filename. So video filenames would change from showing first motion detect time to the video start time with one second resolution. I think this is better and the pre capture time is known so time of detect info is not lost.

Then I can add a new variable option ($F) to the video filename format so that the fractional second part of the first video iframe time can be embedded in the filename. This fraction would come from the GPU iframe time stamp so is the time embedded in the h264 video data and should be as accurate as we can get.

So, the default now in pikrellcam.conf is that the video filename format shows sequence numbers with:

Code: Select all

video_motion_name_format motion_%F_%H.%M.%S_$N.mp4
and if I make these changes, that could be changed to:

Code: Select all

video_motion_name_format motion_%F_%H.%M.%S_$F.mp4
to get the filename to show the actual video start time with say 0.01 fractional second resolution. You could then parse the filenames to get accurate start times. This time should be in sync with the time annotation you see overlayed on the video and is not affected by any pikrellcam system time sampling.

Then offsets into the video could be calculated based on frame rate.
But jbeale points out that actual framerates can vary from what's configured so maybe there can be an actual framerate calculated and presented somehow. I'll have to think about how to do that, maybe in /run/pikrellcam/state.

I've got some weekend plans, but if this could get you something useful I could do something about it next week.

User avatar
jbeale
Posts: 3499
Joined: Tue Nov 22, 2011 11:51 pm
Contact: Website

Re: PiKrellCam: motion vector detect + OSD web interface

Fri May 06, 2016 7:08 pm

billw wrote:...to get the filename to show the actual video start time with say 0.01 fractional second resolution. You could then parse the filenames to get accurate start times. This time should be in sync with the time annotation you see overlayed on the video and is not affected by any pikrellcam system time sampling.
If the accuracy relative to "real time" was anything like such 10 msec precision, that would be quite impressive. I think NTP does keep my Pi's system time with stability of that order, but I don't know about OS latencies when you try to timestamp things. If you use GPU timestamps and somehow accurately know the offset between GPU and system time as aligned via NTP, I guess that works. At any rate, I would be happy to test things...

billw
Posts: 403
Joined: Tue Sep 18, 2012 8:23 pm

Re: PiKrellCam: motion vector detect + OSD web interface

Fri May 06, 2016 7:29 pm

jbeale wrote: If the accuracy relative to "real time" was anything like such 10 msec precision, that would be quite impressive.
Yeah, it did occur to me it might be safer to say something like 0.1 sec resolution which for a video start time should be good enough.

rpiatf2f10
Posts: 27
Joined: Sat Apr 23, 2016 10:16 pm

Re: PiKrellCam: motion vector detect + OSD web interface

Sat May 07, 2016 11:47 am

Billw,

Once we work out a correct way of mounting to remote CIFS/SMB or NFS as alternative to local SD/USB, it would be nice to have this choice during installation process. It will ask whether we use local SD, or USB, or NFS or CIFS. If it's something we can put into GUI, it would be good as well.

https://github.com/ccrisan/motioneyeos , has a nice way of getting CIFS working. It might be something we can learn from it.

Thanks in advance.

User avatar
Gavinmc42
Posts: 3916
Joined: Wed Aug 28, 2013 3:31 am

Re: PiKrellCam: motion vector detect + OSD web interface

Sat May 07, 2016 1:38 pm

Tripped over this while looking at how picamera handles motion.
Network triggered cameras
http://compoundpi.readthedocs.io/en/release-0.4/#
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
jbeale
Posts: 3499
Joined: Tue Nov 22, 2011 11:51 pm
Contact: Website

Re: PiKrellCam: motion vector detect + OSD web interface

Sat May 07, 2016 6:00 pm

If you want to trigger several different cameras at once over the network, my quick test suggests you can get synch (at least, local timestamp agreement) to about 10 msec. Below screenshot shows PuTTY sessions to six different RPi boxes, all on the LAN. One is sending out a broadcast UDP and all six of them receive it with timestamp difference within a 10 msec window. The local machine (rp21) reports it first, as you might expect, about 5 msec earlier. The 4 boxes on the same hub (rp23, rp27, rp30, rp31) report timing within 2 msec of each other. The rp6 machine is last as it connects through two different hubs.

Image

broadcast UDP packet send:

Code: Select all

echo -n "TEST" | socat - udp-datagram:255.255.255.255:55555,broadcast
broadcast UDP packet receiving code, in Python:

Code: Select all

#!/usr/bin/python
import select, time
from socket import *

port = 55555 # port to receive broadcast message
bufferSize = 1024  # maximum message size

s=socket(AF_INET, SOCK_DGRAM)
s.bind(('<broadcast>',port)) # connect localhost to this port
s.setblocking(0)

while True:
    result = select.select([s],[],[])
    msg = result[0][0].recv(bufferSize)
    print msg + " " + ("%.6f" % time.time())

jefferyanderson
Posts: 4
Joined: Thu May 05, 2016 12:01 pm

Re: PiKrellCam: motion vector detect + OSD web interface

Sun May 08, 2016 2:39 am

to get the filename to show the actual video start time with say 0.01 fractional second resolution. You could then parse the filenames to get accurate start times. This time should be in sync with the time annotation you see overlayed on the video and is not affected by any pikrellcam system time sampling.

Then offsets into the video could be calculated based on frame rate.
But jbeale points out that actual framerates can vary from what's configured so maybe there can be an actual framerate calculated and presented somehow. I'll have to think about how to do that, maybe in /run/pikrellcam/state.

I've got some weekend plans, but if this could get you something useful I could do something about it next week.
Bill,
That sounds great, parsing the time from the filename was what I had originally planned anyway, so that would be fantastic. If this is something you could look at in the next couple of weeks that would be fantastic. 100ms would be perfectly fine for my application.
Best Regards,
Jeff

rpiatf2f10
Posts: 27
Joined: Sat Apr 23, 2016 10:16 pm

Re: PiKrellCam: motion vector detect + OSD web interface

Mon May 09, 2016 11:33 am

HI Billw and all,

FYI, I tried using 100M RAM (my pi is PI B+) as alternative media folder, like this,
tmpfs /mnt/ram tmpfs defaults,noatime,nosuid,mode=0755,size=100m 0 0
and
media_dir /mnt/ram/media
It worked well, with root:root to /mnt/ram and pi:www-data to /mnt/ram/media.

In the future, it might be nice to have option of using some available RAM, especially for Pi2/3, to record and store these. We currently use RAM for temporary file anyway. Then, the follow-up script can have options of either move it to local SD/USB or move it to remote SSHFS/curlftps/FTP/SCP. Or, other options would be leverage NFS/CIFS for write/store to those places, or move/copy it to those places. Once it's moved, files can be deleted from PI.

In the web view window, we can have a view of vids/pics on RAM (if we set up moving file only previous day's, not today's) and view of vids/pics already moved to SD/USB, SSHFS/culftps/FTP/SCP, or NFS/CIFS. In the configuration ,we can defined these places, and PI can access them and get those files.

Just some thoughts to share .

thanks

rpiatf2f10
Posts: 27
Joined: Sat Apr 23, 2016 10:16 pm

Re: PiKrellCam: motion vector detect + OSD web interface

Mon May 09, 2016 12:36 pm

HI Guys,

Jasimo kindly shared his notes of getting NFS working as "media_dir", instead of SD card. However,
I tried to follow up his guide. The only problem now is that my NAS (OpenMediaVault) is doing NFS V4 and I can not mount share with "v3" option.

I can mount it without "v3"; however, the directory will be
drwxrwsrwx 3 4294967294 4294967294 786432 May 8 11:46 cam
drwxr-sr-x 2 4294967294 4294967294 4096 May 8 19:48 media
[email protected]:/mnt/cam/media $ touch test5.txt
touch: cannot touch ‘test5.txt’: Permission denied

So, I don't have permission within "media" subfolder. However, I can use "sudo" to create file.

FYI, I tried using 100M RAM as alternative media folder, like this,
tmpfs /mnt/ram tmpfs defaults,noatime,nosuid,mode=0755,size=100m 0 0
and
media_dir /mnt/ram/media
It worked well, with root:root to /mnt/ram and pi:www-data to /mnt/ram/media.

This will likely be the reason. Any ideas to get around this? Anyone has some tips of getting around this NFS4 issue?

Here's the settings on NAS server:
[email protected]:~# cat /etc/exports
# /etc/exports: the access control list for filesystems which may be exported
# to NFS clients. See exports(5).
/export/Recording 192.168.0.0/16(rw,sync,no_subtree_check,secure)

# NFSv4 - pseudo filesystem root
/export 192.168.0.0/16(ro,fsid=0,root_squash,no_subtree_check,hide)

[email protected]:~# cat /etc/idmapd.conf
[General]

Verbosity = 0
Pipefs-Directory = /var/lib/nfs/rpc_pipefs
# set your own domain here, if id differs from FQDN minus hostname
#Domain = localdomain

[Mapping]

Nobody-User = nobody
Nobody-Group = nogroup


Thanks in advance.


******* Jasimo's shared notes of setting up NFS as media folder*********
First I have done the following:

Code: Select all
sudo aptitude install nfs-common portmap
sudo service rpcbind start
sudo update-rc.d rpcbind enable
sudo reboot


after this NFS mappings are possible, then i made:

Code: Select all
sudo mkdir /mnt/cam

to create a folder under /mnt

I got problems to map the nfs drive in fstab, therefore I add the following at the end of the file /etc/rc.local just before "exit 0
" like this:

Code: Select all
sudo nano /etc/rc.local

add

Code: Select all
sleep 5
sudo mount -t nfs -o v3,rw,soft,nolock,wsize=8192,rsize=16384 IP_OF_YOUR_NAS:/NAS_SHARE /mnt/cam
exit 0

after this do a reboot and check if you can reach the nas under /mnt/cam
If you can reach your nas under /mnt/cam edit pikrellcam.conf parameter to

Code: Select all
media_dir /mnt/cam/media
****************

User avatar
Gavinmc42
Posts: 3916
Joined: Wed Aug 28, 2013 3:31 am

Re: PiKrellCam: motion vector detect + OSD web interface

Wed May 11, 2016 4:48 am

I currently get 1500 triggers a day from passing cars, per camera, each video is about 1-3MB.
Now have 4 cameras going, total 12GB per day.

I need to solve the network drive and also reject the passing cars.
By using a RAM drive then checking the vector csv I can delete all the fast cars and only grab the slow and parking ones.

I really love the snapshot thumbnails, but it is a pain checking 1500/night. :roll:
Miss a day copying files and the SD or USB drive fills up.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
jbeale
Posts: 3499
Joined: Tue Nov 22, 2011 11:51 pm
Contact: Website

Re: PiKrellCam: motion vector detect + OSD web interface

Wed May 11, 2016 5:22 pm

Certainly you want a fully automated system; scripts should be transferring and deleting files from local storage, not manual! More local storage is possible, eg. 32, 64 or 128 GB USB flash drives. I have a Chromebook reflashed to Linux for my central storage server, uses 1TB and 2TB USB3 HDDs for medium-term storage. I don't know your situation but I'd aim for at least 1 week of archives. If there's an issue in the neighborhood you only hear about a few days later (or you're on vacation), you may still have something useful. In my area the common issues that come up are lost pets and lost elderly relatives, besides the occasional break-in.

It's impressive that PiKrellCam can reliably trigger even on rather small dogs on the opposite side of the street, which other motion-detectors could not do.

User avatar
Gavinmc42
Posts: 3916
Joined: Wed Aug 28, 2013 3:31 am

Re: PiKrellCam: motion vector detect + OSD web interface

Thu May 12, 2016 4:37 am

Yep Pikrellcam is very impressive, three weeks in and I am still amazed at how good it is.
And how fast it fills up my drives:)

Triggered on a bird, zooming across the frame 30+m away.
We think is was a bird, looks more like a flying Playstation controller:), Micro UFO?
Need to get my 8Mps swapped in and focused.

Playing with this software, it looks like focusing V2's at more than 5m will do infinite focus.
http://www.dofmaster.com/dofjs.html

Wish I could get some PiDrives, only UK?USA?
Had one working with a 1TB USB drive, no problem, but I only have one.
Thinking my 8Mp V2s will go on my Pi2's and make a big ram drive for temp storage.
Then check the vector csv to only copy the slow stuff.
Even with pre and post capture set to minimum, 16GB USB sticks fill up fast.

I could have one setup for detection and it then triggers the others to start recording.
Or spend more time trying to hack the motion.c file to reject the fast moving cars.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

BerryPicker
Posts: 177
Joined: Tue Oct 16, 2012 3:03 pm
Location: The East of England

Re: PiKrellCam: motion vector detect + OSD web interface

Fri May 13, 2016 2:22 pm

jbeale wrote:
Gavinmc42 wrote:Tried again with Jessie light, still got issues, Raspbian works but wastes 4GB.
What were the issues? I just installed on Jessie Light and it worked for me without issues. The only difference was I installed 'git' first because it wasn't there by default. It's nice to start with an 8 GB card only 18% used by the system, instead of over 50% used.

Code: Select all

cd /home/pi
sudo apt-get install git
git clone https://github.com/billw2/pikrellcam.git
cd pikrellcam
./install-pikrellcam.sh
Alas, given the apparent simplicity, I feel ashamed at having no success. Spotting there is now a May jessie light, and with a new 16GB card delivered, I was keen to attempt an install. All went well without errors until after the processing of triggers for man-db. Then I spotted, as they flew by, reports of uninitialised values, and then dependency problems - leaving unconfigured. On my first attempt I had asked for start on boot, but this resulted in a frozen pi. So I gave the card a fresh image and tried another install without start on boot. The reported errors were the same. The pi did not freeze, but on reboot reported fail to start (web and php). If others are having success installing on the May jessie light, please say to give me eccouragement to investigate the problem with my install.

rpiatf2f10
Posts: 27
Joined: Sat Apr 23, 2016 10:16 pm

Re: PiKrellCam: motion vector detect + OSD web interface

Fri May 13, 2016 3:40 pm

HI,
To encourage a bit .... I have been using Jessie Light with no issues.

What I did is to just have GIT installed and then, pulled script to get Pikrellcam installled and it worked after following it's instruction.

Not sure I can help much, but an encouragement.
My pi is pi B+.

BerryPicker
Posts: 177
Joined: Tue Oct 16, 2012 3:03 pm
Location: The East of England

Re: PiKrellCam: motion vector detect + OSD web interface

Fri May 13, 2016 4:58 pm

rpiatf2f10 wrote:HI,
To encourage a bit .... I have been using Jessie Light with no issues.
Thanks for that confirmation. I now also tried with my B+ :? No luck, but I suspect the memory card, as errors are now showing just installing git! My first for a brand new card failure :(

Edit: It seems that certain Kingston cards are incompatable with the May Jessie Light image
viewtopic.php?p=977997#p977997
Last edited by BerryPicker on Thu May 19, 2016 7:48 am, edited 1 time in total.

User avatar
Gavinmc42
Posts: 3916
Joined: Wed Aug 28, 2013 3:31 am

Re: PiKrellCam: motion vector detect + OSD web interface

Sat May 14, 2016 4:01 am

Got one Jessie and 3 Jessie Lites running.
Bad cards can stuff things up, had two fail in the last few weeks, but they are getting to 4yrs old.
I don't trust 16GB cards yet, dollar wise 16, 32 are ripe for counterfeiting.
Only have one 16GB SD and one 16GB micro SD, mostly for reliability testing.

Going to try the new Lite on 4GB SD with 16GB USB drives this weekend.
Most of my 8GB SD's are only class 4, my 4GB are class 6.
Been checking websites for industrial quality SD cards.
Got two old 128MB SD's that must be 10+yrs old and they just keeps on working.
But I run piCore OS on them, it runs from ram and does not write back to the card unless told too.

Power supply and SD card are the two main problems to look at if things go pear shaped.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
Gavinmc42
Posts: 3916
Joined: Wed Aug 28, 2013 3:31 am

Re: PiKrellCam: motion vector detect + OSD web interface

Sat May 14, 2016 4:20 am

Just been reading up.
Another option is boot from SD and put the rest on an USB drive.
The new Jessie has some backup tools now, might be useful?
Boot from SD then load OS from a PiDrive, reminds me of how PC's work :ugeek:
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
Gavinmc42
Posts: 3916
Joined: Wed Aug 28, 2013 3:31 am

Re: PiKrellCam: motion vector detect + OSD web interface

Sat May 14, 2016 8:19 am

Having trouble with the latest Jessie Lite and 4.4.10 kernel.
Not sure yet why, too many issues.
dpkg: warning: parsing file '/var/lib/dpkg/status' near line 13311 package 'libvpx1':
missing description
dpkg: warning: parsing file '/var/lib/dpkg/status' near line 13311 package 'libvpx1':
missing maintainer
dpkg: error: parsing file '/var/lib/dpkg/status' near line 13311 package 'libvpx1':
E: Sub-process /usr/bin/dpkg returned an error code (2)
Adding a pikrellcam autostart command to /etc/rc.local:
su pi -c '(sleep 5; /home/pi/pikrellcam/pikrellcam) &'
Adding to /etc/sudoers.d: www-data permission to run pikrellcam as user pi:
cp: cannot stat ‘etc/pikrellcam.sudoers’: No such file or directory
sed: can't read /tmp/pikrellcam.sudoers.tmp: No such file or directory
sed: can't read /tmp/pikrellcam.sudoers.tmp: No such file or directory
chown: cannot access ‘/tmp/pikrellcam.sudoers.tmp’: No such file or directory
chmod: cannot access ‘/tmp/pikrellcam.sudoers.tmp’: No such file or directory
mv: cannot stat ‘/tmp/pikrellcam.sudoers.tmp’: No such file or directory
Turning off nginx access_log.
Installing /etc/nginx/sites-available/pikrellcam
nginx web server port: 80
nginx web server root: /home/pi/pikrellcam/www
cp: cannot stat ‘etc/nginx-jessie-site-default’: No such file or directory
sed: can't read /etc/nginx/sites-available/pikrellcam: No such file or directory
Job for nginx.service failed. See 'systemctl status nginx.service' and 'journalctl -xn' for details.
cp: cannot stat ‘*’: No such file or directory
Could be SD card issues too, getting fcsk messages.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
Gavinmc42
Posts: 3916
Joined: Wed Aug 28, 2013 3:31 am

Re: PiKrellCam: motion vector detect + OSD web interface

Sat May 14, 2016 8:33 am

Ok, almost nothing got installed.
/www seems to be the only folder with stuff in it.
nothing in /src, /scripts, etc

Done this twice, not getting fixed.
./install-pikrellcam.sh
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

User avatar
Gavinmc42
Posts: 3916
Joined: Wed Aug 28, 2013 3:31 am

Re: PiKrellCam: motion vector detect + OSD web interface

Sat May 14, 2016 8:47 am

Hmm, think I may have found the problem.
viewtopic.php?f=63&t=148064
Don't use rpi-update?

Having trouble writing images to 6 micro SDs.
All past use by date?
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

BerryPicker
Posts: 177
Joined: Tue Oct 16, 2012 3:03 pm
Location: The East of England

Re: PiKrellCam: motion vector detect + OSD web interface

Sat May 14, 2016 10:21 am

Gavinmc42 wrote:Bad cards can stuff things up ...
And this was my problem. Whereas a 16GB Class 10 Kingston gave errors :( , an 8GB Class 4 didn't. :) A 16GB class 10 SanDisc is also up to the task. :P Oddly, an fsck of the dodgy 16GB class 10 Kingston reports zero errors. :?
Back on topic, I thank billw for releasing an excellent upgrade, with the new capabilities well thought out and explained.

Return to “Camera board”