Current Server Time: 10:04 (Central Europe)

#apertus IRC Channel Logs

2018/09/20

Timezone: UTC


01:00
philippej
left the channel
01:00
Bertl_oO
left the channel
01:00
Bertl_oO
joined the channel
01:00
philippej|away
joined the channel
01:01
philippej|away
changed nick to: philippej
01:12
mauz555
joined the channel
01:12
mauz555
left the channel
02:05
notbrent19
joined the channel
02:10
notbrent19
left the channel
04:03
futarisIRCcloud
joined the channel
04:54
BAndiT1983|away
changed nick to: BAndiT1983
05:04
BAndiT1983
changed nick to: BAndiT1983|away
05:04
BAndiT1983|away
changed nick to: BAndiT1983
07:07
se6astian|away
changed nick to: se6astian
07:11
defkult
joined the channel
07:14
defkult
left the channel
07:28
wiiguy16
joined the channel
07:30
wiiguy16
left the channel
07:34
phantom9412
joined the channel
07:35
phantom9412
left the channel
08:43
futarisIRCcloud
left the channel
09:24
skade5
joined the channel
09:24
FastJack2
joined the channel
09:25
FastJack2
left the channel
09:30
skade5
left the channel
09:47
decay_
joined the channel
09:51
decay_
left the channel
09:54
TheSchaf
joined the channel
09:55
TheSchaf
left the channel
10:01
halvor36
joined the channel
10:01
halvor36
left the channel
11:27
hiya19
joined the channel
11:30
hiya19
left the channel
11:31
m0zes16
joined the channel
11:33
m0zes16
left the channel
11:47
AimHere23
joined the channel
11:47
AimHere23
left the channel
11:50
EchoPapa
joined the channel
11:53
EchoPapa
left the channel
11:58
aombk
left the channel
12:43
LordVan
joined the channel
13:42
niteesh
joined the channel
14:08
BAndiT1983
changed nick to: BAndiT1983|away
14:09
aombk
joined the channel
14:13
LordVan
left the channel
14:42
stevenbell
joined the channel
14:56
se6astian
14:56
se6astian
hi stevenbell
14:56
se6astian
14:56
se6astian
14:56
se6astian
14:56
se6astian
14:56
se6astian
14:56
se6astian
14:56
se6astian
14:56
se6astian
14:57
firemouth
joined the channel
14:59
firemouth
left the channel
15:00
stevenbell
hi se6astian!
15:02
GedeonNyengele
joined the channel
15:02
se6astian
as I just read via email Bertl_oO will only be here in 90 minutes
15:02
se6astian
15:03
stevenbell
I saw that; I can hang around for a bit, but I don't know about Gedeon and Keyi.
15:04
GedeonNyengele
I won't be around then. I will have to go to work.
15:04
se6astian
right
15:04
se6astian
lets chat now and Bertl_oO can read the log afterwards
15:05
Keyi
joined the channel
15:05
se6astian
15:05
se6astian
hi here
15:05
se6astian
*there
15:05
Keyi
hello
15:05
stevenbell
GedeonNyengele: Oh, I forgot it's still summer at Stanford. :-)
15:06
Keyi
yep. it's 8 in the morning here
15:06
se6astian
should we go to a private channel or is it OK if we stay in the public one here?
15:07
stevenbell
doesn't matter to me; I assume it's normally quiet enough?
15:08
stevenbell
and it's fine with me if someone else wanders into the conversation.
15:08
se6astian
good then lets stay here
15:09
se6astian
there are public logs, just so you are aware
15:09
se6astian
no secrets here :)
15:10
Keyi
cool. do you have any agenda for this meeting?
15:10
se6astian
right then many thanks for taking the time to come here, the original spark for making this meeting happen my and herberts the planned silicon valley visit mid october for the google summer of code mentor summit and so I contacted steven if he was still at stanford and if it would make sense to meet
15:10
Gedeon
joined the channel
15:11
se6astian
15:11
se6astian
so the main agenda is just to establish first contact and see if there is anything in our and your work that would make sense to collaborate or discuss
15:12
Keyi
sure. Steven do you want to briefly discuss your f4graph pipeline?
15:12
stevenbell
yeah, I have two points -- one is to share some of the stuff we've been doing at Stanford with cameras/FPGAs and see if some of it might be useful to you, and second to see if there are small projects that my students here at Tufts might be able to work on.
15:13
se6astian
great, I fear now we would have really needed Bertl_oO :)
15:14
stevenbell
maybe so. I can always follow up with him later...
15:14
se6astian
stevenbell: for your students are you looking for small software projects, or hardware?
15:14
se6astian
because our google summer of code idea page would be a perfect fit I think: https://lab.apertus.org/project/view/20/
15:14
se6astian
if the focus is similar
15:15
se6astian
for GSoC its meant to be a 3 month full time project but we try to always use projects that scale well depending on how he student is progressing
15:15
se6astian
so a basic goal/project can be much smaller/simpler
15:15
stevenbell
I have a couple possible cases: one is a student that wants to do a research project over the summer, and there a GSoC project is probably the right size.
15:16
se6astian
perfect
15:16
se6astian
one is to share some of the stuff we've been doing at Stanford with cameras/FPGAs and see if some of it might be useful to you <- sounds great, please do :)
15:16
se6astian
and the f4graph pipeline
15:16
stevenbell
the other is a team of students wanting to do a course project; I'm currently teaching a course about cameras and image processing. http://www.ece.tufts.edu/ee/193HIP/
15:17
stevenbell
that's generally smaller scope, and likely to be lower quality, but if there are "go investigate this and see what's possible" kind of projects, those might also be appropriate.
15:17
Keyi
also just FYI we're designing a new CGRA chip that's specialized for image processing. we just passed our first milestone to run harris corner end-to-end
15:19
stevenbell
so one thing we have working fairly well is a system to build image processing pipelines from the Halide language.
15:20
stevenbell
I saw there have been a couple of GSoC projects to build HLS or VHDL designs to do various image processing tasks, and the Halide->FPGA system makes that way faster.
15:21
stevenbell
We're typically able to develop code from scratch and have it running on the hardware within a couple days (rather than a few weeks), and the hardware result is competitive with hand-written HLS.
15:21
stevenbell
("result" being the throughput and resource utilization on the FPGA)
15:21
se6astian
very interesting!
15:23
stevenbell
The repository is here: https://github.com/jingpu/Halide-HLS
15:24
stevenbell
if bertl_oO or anyone else wants to give it a try, we'd be happy to help. It's research code, so there are definitely rough edges, but it mostly works. :-)
15:24
se6astian
right
15:25
stevenbell
here's a link to the paper about it: https://arxiv.org/abs/1610.09405?
15:27
se6astian
thanks, Bertl_oO will surely check it out
15:27
se6astian
anything else that we could talk about now?
15:28
stevenbell
I can give a quick overview of f4graph, which is my thesis work (with Gedeon and Keyi)
15:29
se6astian
please do
15:29
stevenbell
the basic idea is to provide an API for controlling multiple cameras (and other peripherals) within a camera system.
15:31
stevenbell
I don't know how true this is for your current applications, but in a lot of computational photography apps, it's important to have flexible and precise control of when actions happen in the system.
15:31
stevenbell
So, take a shot on camera 1, and 5ms later take a shot on camera 2.
15:31
stevenbell
Or, expose for 25ms and fire the flash for the first 3ms of the exposure.
15:33
stevenbell
And these kinds of things aren't really possible within current camera frameworks on Android/iOS.
15:33
BAndiT1983|away
changed nick to: BAndiT1983
15:33
se6astian
right, for the film making applications that isnt required, but since the AXIOM Beta runs linux inside you can very easily created bash, python, etc. scripts that do something like this
15:33
se6astian
all camera parameters can be controlled from the commandline
15:34
se6astian
there is no nice universal API yet though
15:34
stevenbell
hmm, I wonder if it would be a good candidate for an implementation of FCam and/or f4graph?
15:35
stevenbell
I think you're familiar with the Frankencamera project?
15:35
se6astian
BAndiT1983: perfect timing, he is working on the camera internal control daemon
15:35
se6astian
yes, I followed the progress with the different stages of the frankencamera pretty much from the beginning I think
15:36
stevenbell
One of their problems was that the hardware wasn't always able to do what the FCam API allowed; the result is that although Android basically implemented FCam, it had to be "dumbed down" to match what commodity hardware could realistically provide.
15:37
stevenbell
but with your camera, anything is possible. :)
15:37
se6astian
in a nutshell yes :)
15:38
stevenbell
Gedeon and Keyi, do you want to chime in about f4graph or other stuff?
15:39
Gedeon
stevenbell: You pretty much covered it
15:40
Keyi
sure. I was reading through the gosc tasks and it seems like it overlaps a lot of stuff we did with f4graph
15:40
se6astian
great
15:40
mauz555
joined the channel
15:40
Cprossu4
joined the channel
15:41
Keyi
we have implemented an zero-copy linux kernel driver that interfaces with our camera pipeline
15:41
stevenbell
yeah, here's the repository: https://github.com/stevenbell/ultrazynqbuilder
15:42
se6astian
the basic idea of the camera internal control daemon ( BAndiT1983 please chime in) is that it allows adding new "plugins" for any kind of API in the future
15:42
Cprossu4
left the channel
15:42
BAndiT1983
am here
15:43
Gedeon
if you don't mind me asking, what exactly does this camera daemon do?
15:43
se6astian
its the "bridge" between internal camera parameters and the outside world
15:43
sebix
joined the channel
15:43
sebix
left the channel
15:43
sebix
joined the channel
15:43
se6astian
get/set parameters of any kind (image sensor related, image processing related, etc.)
15:44
Gedeon
so basically a driver to expose the camera's register space (configuration space)?
15:44
BAndiT1983
there are different sets of configs or sensors, so register space is just one of them
15:46
Gedeon
so these configs aren't accessed through an address space?
15:46
Keyi
how fast can you change these settings? what's the granularity of the changes?
15:47
BAndiT1983
for image sensor it's done through memory mapped space, other stuff is planned to be accessed through IO pipes or whatever is present in linux
15:47
BAndiT1983
general communication is done through websockets server, there is also a command line tool, to set things directly, as a replacement for scripts
15:48
BAndiT1983
because of gsoc, the development stalled a bit, but should resume soon, when some other apertus tasks are reviewed and planned
15:48
stevenbell
sure, makes sense. Can you share a pointer to the code?
15:48
BAndiT1983
settings should change instantly, this also depends on how camera internal electronics processes them, granularity is variable
15:49
BAndiT1983
https://github.com/apertus-open-source-cinema/beta-software/tree/master/software/control_daemon
15:49
stevenbell
se6astian, where are you at with production? I saw you're in a cool new facility. If I had students who wanted to experiment with a camera next summer, would it be possible to buy one?
15:50
se6astian
yes
15:50
weylin7
joined the channel
15:50
weylin7
left the channel
15:50
se6astian
we are producing devloper kits in small quantities currently
15:51
se6astian
we also have a camera at the office with remote access for students/developers
15:52
stevenbell
oh, cool! you just ssh in?
15:52
se6astian
that https://wiki.apertus.org/index.php/AXIOM_Beta_Remote_Access
15:52
se6astian
stevenbell: yes, you can even remotely turn lights on and off :)
15:53
se6astian
We are thinking about further improving this setup (like putting the whole setup into a dark box - making environment 100% controlable
15:53
se6astian
)
15:53
stevenbell
oh, that's even better than I expected. :-) We still usually have to go manually reboot our boards when we crash them.
15:54
jucar
joined the channel
15:55
m6w616
joined the channel
15:56
stevenbell
I'd better sign off soon, and I imagine Gedeon needs to get to work. Anyone have other questions/comments?
15:56
se6astian
I think we are done
15:56
Keyi
I have some quick questions
15:57
Keyi
what type of FPGA are you guys using?
15:57
stevenbell
it's the microzed, right?
15:58
se6astian
yes
15:58
se6astian
xilinx zynq 7020
15:58
stevenbell
yeah, we don't have any microzeds, but we have a pile of 7020 boards sitting around.
15:59
Keyi
okay thanks. I'm working on fast placement on fpga and plan to interface vivado very soon
16:00
se6astian
your welcome
16:00
m6w616
left the channel
16:00
se6astian
great meeting you
16:00
Gedeon
I think that if we were to somehow get f4graph and this camera system to properly work, we'll have a one-of-a-kind product
16:01
se6astian
indeed
16:01
se6astian
lets continue via email and schedule another chat if it makes sense
16:01
Gedeon
definitely
16:01
Keyi
sure. thank you very much
16:01
stevenbell
and I'll be on the lookout for any students who might want to make that happen.
16:01
stevenbell
thanks all for taking time to chat!
16:02
se6astian
all the best
16:02
se6astian
I am off too
16:02
se6astian
bye
16:02
se6astian
changed nick to: se6astian|away
16:02
Gedeon
left the channel
16:02
stevenbell
left the channel
16:04
Keyi
left the channel
16:06
GedeonNyengele
left the channel
16:10
jucar
left the channel
16:11
sebix
left the channel
16:31
mauz555
left the channel
16:49
bwoah17
joined the channel
16:51
Bertl_oO
I'm around now ...
16:52
bwoah17
left the channel
16:54
BAndiT1983
changed nick to: BAndiT1983|away
16:54
BAndiT1983|away
changed nick to: BAndiT1983
17:08
niteesh
left the channel
17:24
BAndiT1983
changed nick to: BAndiT1983|away
17:50
mauz555
joined the channel
18:23
danieel
left the channel
18:36
se6astian|away
changed nick to: se6astian
18:37
se6astian
changed nick to: se6astian|away
18:37
se6astian|away
changed nick to: se6astian
18:38
se6astian
changed nick to: se6astian|away
18:57
niemand
joined the channel
18:57
niemand
left the channel
18:57
niemand
joined the channel
19:05
mauz555
left the channel
19:06
BAndiT1983|away
changed nick to: BAndiT1983
19:10
se6astian|away
changed nick to: se6astian
19:14
se6astian
left the channel
19:14
RexOrCine|away
left the channel
19:14
BAndiT1983
left the channel
19:14
philippej
left the channel
19:14
BAndiT1983|away
joined the channel
19:14
RexOrCine|away
joined the channel
19:14
se6astian|away
joined the channel
19:14
se6astian|away
changed nick to: se6astian
19:14
BAndiT1983|away
changed nick to: BAndiT1983
19:14
philippej|away
joined the channel
19:14
philippej|away
changed nick to: philippej
19:19
ChanServ
19:19
se6astian
19:19
se6astian
19:19
se6astian
19:19
se6astian
19:34
danieel
joined the channel
20:03
opi20
joined the channel
20:07
opi20
left the channel
20:21
squarebracket16
joined the channel
20:23
squarebracket16
left the channel
20:58
se6astian
changed nick to: se6astian|away
21:01
niemand
left the channel
21:23
BAndiT1983
changed nick to: BAndiT1983|away
21:23
BAndiT1983|away
changed nick to: BAndiT1983
21:23
hexa2k910
joined the channel
21:28
hexa2k910
left the channel
21:44
BAndiT1983
changed nick to: BAndiT1983|away
21:52
danieel
left the channel
22:25
danieel
joined the channel
22:51
colinbits25
joined the channel
22:51
colinbits25
left the channel
23:32
danieel
left the channel
23:51
mauz555
joined the channel
23:59
futarisIRCcloud
joined the channel