Fast MD with burst/preview mode

mx3

slchdk
OK, well I tried increasing the rows/cols, and also tried decreasing threshold, and also tried decreasing pixel step, and tried increasing the compare interval, and my A650 won't go lower than 120 ms (and still gets 200 ms frequently).

MD uses "vid_get_viewport_live_fb" function to get data for processing.
Some ports have "standard" function which uses only one of three available LCD buffers to do math processing.
Other ports have "advanced" version of this function which uses all three buffers.
I just looked into build 638 and it seems port for A650 uses "standard" version of this function.
it very well explains chaotic trigger time - It can be improved to give more smaller/stable/predictable reaction times.

-----------------
edit:
If you want to participate in solving this puzzle (you will need to setup compiler and cook CHDK yourself) you have two options:
1) tough way - digging into ROM dump using IDA to find out addresses (can take many-many-many hours or you can ask some guru).
2) dumb way - dumping RAM and do some comparing - I provided tools somewhere on this forum. This way do not give 100% guarantee but it is possible you can do it in no-time.... 1-3 hours:
- take several dumps (it takes 1-30min)
- analize generated output (it can take 5-30 min)
- change sources of CHDK and cook new build - (5-30 min)

slchdk

mx3: I checked the wiki and did not see anything on how to set up a dev environment or check out the code. Is there a thread that can help me get started? I might have some time later, but I'm sure I'd need some help with this.

reyalp

slchdk
mx3: I checked the wiki and did not see anything on how to set up a dev environment or check out the code. Is there a thread that can help me get started? I might have some time later, but I'm sure I'd need some help with this.

http://chdk.wikia.com/wiki/Modifying_the_CHDK_Sources

I'd suggest read the other pages linked from http://chdk.wikia.com/wiki/For_Developers as well

mx3

slchdk
Is there a thread that can help me get started?

my lazy method I mentioned (see attached file)
and look into other ports lib.c for example of "advanced" version of this function

slchdk

OK, I got the CHDK-Shell going (great job whim!).

mx3: looks like I need to get 2 dump files before I can run the stuff in the .rar file from the other thread you mentioned. I assume it's the same kind of dump file mentioned here? http://chdk.wikia.com/wiki/Porting_the_CHDK#Q._How_can_I_get_a_firmware_dump.3F

If so, then I'll just use CardTricks to create a universal dumper card for DryOS. What are those Strings and Clean buttons in CardTricks next to "After Dumping-->" ?

Then I put the SD card into my A650 and just press the shutter button and a dump file will be created?

Thanks.

mx3

those dumps you refers to are ROM(code) dumps.
you need to make RAM dumps to find out where camera software stores LCD data.
you do not need other tools(cardtricks) to prepare those dumps
to make RAM dumps you have to activate this feature somewhere in ALT debug menu.
then you will be able to make them pressing some button...

fudgey

slchdk
fudgey, great job with this script, and excellent readme file! I noticed in the readme it says to use P mode instead of Auto mode. May I ask why P mode is better than Auto mode? Thank you.

It's been a while since I wrote that, but I think it was because my a570 often crashes with flash. In auto mode (unlike in P, Tv, Av, M modes) it forgets the user's choice of flash mode when powered off and defaults to flash auto (which I think is a good thing for that particular mode, but also a reason to avoid it for MD on a570).

BaGRoS

Can change this:

md_detect_motion a, b, f, n, d, c, 1, t, i, j+1, l+1, a-k, b-m, 9, h, e

md_detect_motion a, b, f, n, d, c, 1, t, i, w, x, y, z, 9, h, e

beffore loop

let w=j+1
let x=l+1
let y=a-k
let z=b-m

slchdk

mx3
those dumps you refers to are ROM(code) dumps.
you need to make RAM dumps to find out where camera software stores LCD data.
you do not need other tools(cardtricks) to prepare those dumps
to make RAM dumps you have to activate this feature somewhere in ALT debug menu.
then you will be able to make them pressing some button...

Hi mx3, I have a couple questions for you, which I posted into the other thread:

https://chdk.setepontos.com/d/405/222

Thanks.

fudgey

BaGRoS
[code]let w=j+1
let x=l+1
let y=a-k
let z=b-m[/code]

Sure, but I can't see much point in doing that. Each line is evaluated during a single tick, which means calculating that over and over again doesn't slow ubasic down. Calculating them at scipt startup like you suggest however slows down script startup by 40 ms, and it wastes 4 variables (not important for a finished script, but it's a pain if you want to extend it -- we only have a-z A-Z variables to use). Btw it's good practice to use capital letters for variables you create inside a script, since input parameters can only be lower case a-z (and this script is already using a large portion of those)... adding new params is not fun at all if all lower case letters are taken already.

Dave/is

On page 11, 5th messeage - May / 2009, 19:04:22 - user Obihoernchen posted photo of some program, which shows delay of the triggering. Can you anyone tell me the name of this program? I want measure delay of triggering of my A590is.
Thanks.

fe50

Dave/is
On page 11, 5th messeage - May / 2009, 19:04:22 - user Obihoernchen posted photo of some program, which shows delay of the triggering. Can you anyone tell me the name of this program? I want measure delay of triggering of my A590is.

Dataghost's webbased MD-meter; you'll find it (and some other apps) here: CHDK wikia: Software - Motion Detection

CanonBlue

Hi, I've been posting about this in a separate thread here but I think that this might be the right place for it, in case anyone else has had the same problem. I downloaded the MD script from this post.

I'm using an SD1100IS / IXUS 80, which means it has no separate video button but rather a slider. I'm using the script for the cameras with no video button.

The motion detection script works very well in photo mode, but it doesn't work in video mode at all. It simply exits immediately when I try to run it. I think that maybe it isn't detecting that the camera is in video mode. Has anyone else had this problem, or know how to fix it? I really want to use the motion detector with video. (Alternately, if anyone knows of another script that would work that would also be great, but I really like how this one works with the photos and I'd love to use it with video).

Thanks!

fudgey

CanonBlue,

First of all, I noticed that in all modes I get a red message in the top left corner of the camera screen that says "uBASIC:9 Parse err". I don't think I've ever noticed that before. Did I do something wrong in copying the script?

Yes, either your script is corrupt (memory card could be corrupt or you may have copied or edited the script incorrectly) or you are running an incompatible CHDK version. Anything that goes wrong for you may be due to this.

I think this matches the error I've been getting from the motion detector script when in video mode. Its difficult to read but I think I have noticed the word exiting. Does that mean that it thinks I'm in play mode when I'm in video mode? Is there some setting that I should change to make it realize its in video mode? I've looked for something like that but couldn't find anything...

Yes. If this is correct, the ixus80_sd1100 get_mode is broken in CHDK. You can probably make things printed by the test script prevail on the screen longer by duplicating each of the if M=? then print "something" lines...say, 20 times each.

(Btw is there a reason that the motion detector script I am using, the first one, doesn't seem to be on the main scripts page? I went there to check and I found a different one instead, but I think the one I've been using works better.)

Because the wiki doesn't allow file attachments and I don't want to read support requests from people who break my scripts incorrectly copy pasting them using weird browsers on even more weird operating systems with increasingly weird character sets. And because I'm downright lazy.

The motion detection script works very well in photo mode, but it doesn't work in video mode at all. It simply exits immediately when I try to run it. I think that maybe it isn't detecting that the camera is in video mode. Has anyone else had this problem, or know how to fix it? I really want to use the motion detector with video. (Alternately, if anyone knows of another script that would work that would also be great, but I really like how this one works with the photos and I'd love to use it with video).

You should try an older version of the MDFB script from here:
https://chdk.setepontos.com/d/471/87. It has a separate still/video setting instead of an automatic mode check. So if video record from scripts works in video mode for your camera by pressing the shutter once for rec and once again later for stop, I believe this older version should work.

CanonBlue

You should try an older version of the MDFB script from here:

I tried that, and it worked, thanks! I haven't tested it extensively, but on my first try it worked to record video. I'll have to experiment with the various settings and stuff later. Thanks for all your help, I'm really glad to have this working!

Because the wiki doesn't allow file attachments and I don't want to read support requests from people who break my scripts incorrectly copy pasting them using weird browsers on even more weird operating systems with increasingly weird character sets. And because I'm downright lazy.

Fair enough, hehe. I only heard about this script because I asked someone on flickr who had posted lightening shots. I'm telling my friends about it.

Yes, either your script is corrupt (memory card could be corrupt or you may have copied or edited the script incorrectly) or you are running an incompatible CHDK version. Anything that goes wrong for you may be due to this.

Anything's possible, but I don't know why my card would be corrupt or the CHDK would be incompatible. I only installed it a couple weeks ago so it should be the most recent version. I did have some trouble installing it but in the end I think I got everything right... but I certainly wouldn't be surprised to find out something else went wrong. Also, as far as copying the script, I used TextEdit (I'm on a mac) and I made sure it was in text only mode and all of that. So I think it was copied correctly, but I don't have a ton of experience with this, so again, who knows... Anyway, it was the only script where I've gotten that uBASIC error, so maybe that points to me copying it wrong? I'll have a look at it again later (I'm rather rushed now with exams and travel coming up) and see if I can figure out the problem.

Yes. If this is correct, the ixus80_sd1100 get_mode is broken in CHDK. You can probably make things printed by the test script prevail on the screen longer by duplicating each of the if M=? then print "something" lines...say, 20 times each.

I'll try that later when I have more time and report back. I *think* that I was getting the same error message in play and video modes, which would point to that being the problem.

tvkas

Hello all, I have been a user of CHDK for over a year and have been getting into it more and more. I have an S5 IS with the latest (as of 3 days ago) build. I got into MD bug after seeing a fantastic lightning storm. I am not that good at code and such but as MX3 said, "If you want to contribute" so I am going to. What might take most of you a few minutes will likely take me several hours/days. But I love CHDK and am bound and determined to learn. I have everything ready, even the compiler shell all setup.

The roadblock I seem to be running into now is simply just getting the dump! I have <ALT> set and press the +/- button and nothing happens. However, if I press <ALT> FUNC it writes a 32+ meg JPG file that I am guessing must be the dump file. Question is, now what do I do? Sorry to be such a neub but we all gotta start somewhere...

tvkas

Nevermind, I found this and I think I will go in that direction for a while to see if I can sort it out....
https://chdk.setepontos.com/d/688/41

gomiboy

Just tried this https://chdk.setepontos.com/d/471/109 on A720IS with the latest build of CHDK and it works a treat 😀

Thanks fudgey

.:olo:.

Nice! Fast MD works well! Small Lightnings Gallery
I wait for the next storm. 😀

Anaglyphic

Hey. I'm sorry to keep this thread going but it's still shorter than the speed test and it's relevant to the general functioning of MDFB...

As you guys know (or at least fudgey) the low-level changes being detected is hard sometimes. I was particularly aggravated in my recent experimenting getting it to trigger on a hue change with equal luma. Seems this is near impossible!?

For example, I'm in NYC, and I'm in the main arriving flight path for LaGuardia Airport. This is a very active airport, I can set my clock by the approach pattern; you can see arrivals stacked up 5 of 6 deep on a clear day, especially at sunset when the nose/wing lights are on. There's also a helipads along the East River and the traffic they generate, as well as pontoon planes that use the river itself...

OK, so on a clear blue sky day, with the screen masked to 1/15th like in the recent speed testing, with pixel step at 0/1 and threshold at 0/1... I have tried every single Channel setting... RGB I wasn't expecting much anyway, but Y & U/V (which I haven't checked the source yet but should be blue-luma and red-luma, respectively.. I'm going under the assumption we are talking about jpeg YUV and not video YUV, although it probably doesn't matter either way.) -- nothing ever triggers!

Now, when a silver aluminium jumbo jet body is only ~3 pixels tall and ~8 pixels long on the viewfinder, it might appear as the same luminance as the background sky, but detection set to Y or U it should have been detected as desaturated from the blue sky, yet it, nor a fleet of those mysterious black helicopters with no markings (conspiracy freaks please Google) are ever enough to trip the capture... even at pixel step 0??

I'm watching the LCD and I can see them clearly, yet they're flying back and forth, all over the active detection area, with not a shot being taken.

Any ideas?

Since this specific MD test is always available, weather permitting, I'd be happy to test any experientially-minded ideas you can dream up. Right up until the time DHS or FBI kick in my door for suspicious activity. :'(

« Previous Page Next Page »