Current Server Time: 20:30 (Central Europe)

#apertus IRC Channel Logs

2013/07/15

Timezone: UTC


11:06
christiankrupa
joined the channel
11:21
Bertl
morning everyone!
11:26
dmj_nova
morning Bertl
11:40
Bertl
hey, I built a stand for my zedboard yesterday (to support it in a tilted position)
11:40
Bertl
http://vserver.13thfloor.at/Stuff/AXIOM/ZedBoardStand.eps
11:40
Bertl
if you have a fablab with a lasercutter in your vicinity (or just a laser cutter) you can use this to make your own
11:41
Bertl
angle and material thickness are adjustable (as well as other parameters)
12:17
dmj_nova
Bertl: oh, very nice
12:18
dmj_nova
I've been thinking this needs that for some time
12:18
Bertl
it's simple but works well
12:18
dmj_nova
actually, a cover might not be a bad idea either
12:18
dmj_nova
and I do have a fablab nearby
12:18
dmj_nova
but open hours aren't until tuesday
12:19
Bertl
I'm not so into covers but I had problems to see the leds (because they are hidden behind the switches :)
12:28
Bertl
you can actually make the design 'click on' by simply adding tiny protrusions? (no idea how to call that in english) to snap into the baseplate
12:32
dmj_nova
tabs?
13:27
Bertl
maybe :)
13:30
Bertl
__ __
13:30
Bertl
| \ / |
13:30
Bertl
| -` '- |
13:30
Bertl
| | | |
13:30
Bertl
| +----------+ |
13:42
dmj_nova
huh?
13:43
dmj_nova
ah, ascii art
13:43
Bertl
that's what I meant in ascii art, yes :)
13:44
dmj_nova
makes more sense with monospaced font :P
13:44
Bertl
who uses proportional fonts on IRC :)
13:45
dmj_nova
pretty sure my irc client has defaulted to Ubuntu
13:45
dmj_nova
(Ubuntu the font)
13:45
Bertl
btw, how is it going with the zedboard setup so far? everything working as planned?
13:45
dmj_nova
I think it's setup, but I haven't had the time to do much with it
13:46
Bertl
real persons (that's political correct) use irssi
13:46
dmj_nova
irssi drives me insane
13:47
dmj_nova
didn't even like it on my phone
13:47
Bertl
hehe, well, I love it
13:47
FergusL
(irssi or weechat)
13:47
Bertl
it is simple (but still powerful) and works (from) everywhere
13:48
Bertl
but hey, it's a choice, not a must .. so if pidgin or xchat or even bitchx is fine for you, so be it
13:48
dmj_nova
yeah, out of everything, pidgin is my preference
13:49
dmj_nova
a few plugins and I'm happy with it
13:49
dmj_nova
xchat also annoyed me, though it wasn't too bad on a phone
13:49
Bertl
yeah, my parents use pidgin as well, and they are happy too
13:50
dmj_nova
loath empathy though
13:50
Bertl
so, are those small protrusions on the ascii art called tabs?
13:51
dmj_nova
it's one word for them
13:51
Bertl
okay, could you describe the entire thing for me?
13:51
Bertl
i.e. what are you seeing
13:52
dmj_nova
I'm seeing two little nubs that hold a thing in place
13:52
dmj_nova
like there's a little overhang on either side
13:53
Bertl
ah, I like knub, that one is actually quite similar to the german slang for it
13:53
Bertl
-k
13:53
Bertl
nub
13:55
Bertl
so, it should be simple to extend the design by adding a few nubs in the right places :) ... maybe I'll do a second version next time which is just click on
13:56
dmj_nova
I've found I like the CLI workflow, but that a bit of chrome can be nice interface-wise for certain tasks.
13:57
dmj_nova
hence why IDEs bug me, but I prefer gedit over vi or nano
13:58
Bertl
yeah, the problem with those IDEs can be seen in the current Xilinx tools
13:58
Bertl
there are 3 different IDEs for the very same task
13:58
dmj_nova
I find I spend more time setting up IDEs than using them
13:58
Bertl
each one can only do part of the job properly, and all of them will be superceeded in the next release (and probably replaced by yet another wizard)
14:00
Bertl
I studied and even taught user interface design, so I'm all for good and useable GUIs, but they are really rare nowadays
14:01
Bertl
and when I read documentation stating that there is no separate tool to do this and that, but you can simply create a fake project to view your file I get really sick :)
14:02
dmj_nova
ugh
14:02
dmj_nova
Bertl: that's cool you taught UI design
14:03
dmj_nova
Bertl: I suspect a large problem with GUIs is that frequently people mistake form and function
14:04
dmj_nova
Or they don't bother with design at all and just put some buttons wherever there's room
14:04
Bertl
a main problem is that there are no good (G)UI design guides for most operating systems and game loaders :)
14:05
Bertl
so folks end up taking whatever they can find and put it together into some abomination called user interface
14:06
dmj_nova
http://cdn.novacut.com/Novacut-12.04-4.jpg
14:07
dmj_nova
http://cdn.novacut.com/novacut-12.03-1.jpg
14:11
Bertl
cool hat!
14:13
dmj_nova
Of course we're not really that concerned with matching the system aesthetic.
14:14
dmj_nova
especially since the standard toolkits really don't offer the right kinds of widgets
14:23
dmj_nova
Bertl: The other problem with GUIs is that it gets very tempting to build your application around the GUI rather than letting the GUI be an interface for your application
14:23
dmj_nova
which means you get stuck to bad UI design
15:15
Bertl
yeah, happens ...
15:15
Bertl
off for a nap .. bbl
15:29
intracube
joined the channel
18:20
dmj_nova
left the channel
19:03
Bertl
back now ...
19:04
Bertl
welcome intracube!
19:05
ApertusWeb1
joined the channel
19:06
ApertusWeb1
changed nick to: Frederir
19:11
Bertl
question to those actually making some movies (or at least planning to make some): wouldn't it be nice to have all kind of motion information from a camaera? I'm thinking about 3-N DOF sensors here, so 3-axis gyroscope, 3-axis acceleration, 3-axis compass and maybe other stuff like air pressure, temperature, humidity, ambient light, etc
19:12
intracube
hi Bertl
19:14
intracube
it would be better to have a separate unit to record that sort of data and sync it via timecode
19:15
intracube
as it'd be an unnecessary expense in a lot of cases
19:15
intracube
but for VFX motion tracking and such, yes, if it could lead to more reliable tracks
19:15
Bertl
guess that would be fine for things like gps, humidity, pressure, etc, but for stuff like gyroscope and acceleration it should definitely be tightly tied to the camera no?
19:18
intracube
still something of a niche feature tho, and it would add significant cost for all users
19:18
intracube
Bertl: what do you mean tightly tied to the camera?
19:18
Bertl
well, a 3-axis accelerometer can be bought for 1-2 USD
19:19
Bertl
a good 3-axis mems gyroscope for 4-6 USD
19:19
intracube
+ development costs for the camera software/firmware
19:20
Bertl
will be almost zero on an open source project, no?
19:20
intracube
+ extra bulk to fit in the camera body (not sure how big these devices area)
19:20
Bertl
tiny
19:22
Bertl
http://www.st.com/web/catalog/sense_power/FM89/SC1448/PF252427
19:22
Bertl
just as example, this one is 4x5mm
19:23
Bertl
(and has 3-axis gyroscope and 3-axis accelerometer)
19:24
Bertl
the cost would be about 7 USD for very low quantities
19:26
Bertl
so back to the original question: would it be useful? or is that some kind of data movie makers/cutters are not really interested in?
19:27
Bertl
(ad tightly tied: I mean inside the camera head? near the actual image sensor)
19:27
Frederir
left the channel
19:30
FergusL
probably for the orientation to check if it lays perfectly flat for example
19:31
FergusL
the acceleration information might be interesting but as is it would require work on the other side
19:32
FergusL
like, in the analysis software doing the actual tracking of video data, one has to code how to use the accel data
19:40
Bertl
of course, but software is flexible in this regard, no?
19:43
FergusL
well, yes !
19:43
FergusL
for the prototype I'm not sure it might be "needed"
19:44
Bertl
not talking about the prototype here
19:44
dmj_nova
joined the channel
19:45
Bertl
just brainstorming here
19:49
FergusL
ok
20:06
intracube
Bertl: question is how accurate are these gyroscopes/accelerometers?
20:06
intracube
I doubt very much they'd be accurate enough by themselves for tracking purposes
20:07
intracube
as we'd be talking about no more than one or two pixels inaccuracy at 2K
20:08
intracube
and thinking about it, could the data really aid existing tracking methods?
20:24
Bertl
no idea if the accuracy is good enough, the mems sensor I picked (by random) has 8.75/1000 degree per second resolution
20:25
Bertl
and 1mg acceleration
20:26
Bertl
but IMHO it would be valuable information to know the camera tilt in all directions, the angular turn rate and the accelleration in each dimension
20:27
Bertl
after all, there is a limit how much you can move/turn without distorting the image sequence, no?
20:50
ApertusWeb2
joined the channel
20:50
ApertusWeb2
changed nick to: frederir
21:16
jucar2
joined the channel
21:18
jucar1
left the channel
22:11
dmj_nova
Bertl: I'd guess it would get off pretty quickly
22:11
dmj_nova
And be a bit wobbly
22:11
frederir
left the channel
22:18
Bertl
well, gravity is always a bit wobbly :) but I guess you are referring to the gyroscope?
22:25
dmj_nova
yeah
22:25
dmj_nova
experience is that they're not that great for absolute precision
22:25
dmj_nova
(at least not the ones in most consumer-grade hardware
22:29
Bertl
not sure that absolute precision is required, but I take it that a 6/9-DOF mems is not something which excites movie folks :)
22:29
dmj_nova
Bertl: That said, I've never tried it
22:30
dmj_nova
Bertl: Actually I meant to say "absolute positioning"
22:31
dmj_nova
aka, I can handhold this camera and get a good track on a 2 minute shot
22:31
Bertl
the interesting part is, that optical image stabilization solely relies on gyroscopes
22:32
dmj_nova
Bertl: That makes sense though
22:33
dmj_nova
Since, optical image stabilization just needs to smooth out relative motions
22:33
dmj_nova
whereas camera tracking requires that if you wander off and come back, that the point will remain the same
22:34
dmj_nova
basically, there's compounding of little errors in tracking, but not in optical stabilization
22:35
Bertl
so how is camera tracking solved atm?
22:36
dmj_nova
basically tracking a set of points in the video as the camera moves
22:37
dmj_nova
and assuming the set of points are fixed relative to each other and knowing the lens you can work it out
22:37
dmj_nova
in software
22:54
Bertl
I see ...