00:23.13 |
Notify |
03BRL-CAD:starseeker * 63361
(brlcad/branches/qtged/src/qbrlcad/cadtreeview.cxx
brlcad/branches/qtged/src/qbrlcad/cadtreeview.h): Using hints from
http://stackoverflow.com/questions/14255224/changing-the-row-background-color-of-a-qtreeview-does-not-work,
get the highlight bars to go all the way across the QTreeView
row. |
00:27.21 |
Notify |
03BRL-CAD:starseeker * 63362
brlcad/branches/qtged/src/qbrlcad/cadtreemodel.cxx: If we've got
the selected entity, maintain the highlighting when closing the
branch |
00:49.44 |
Notify |
03BRL-CAD:starseeker * 63363
(brlcad/branches/qtged/src/qbrlcad/cadtreeview.cxx
brlcad/branches/qtged/src/qbrlcad/cadtreeview.h
brlcad/branches/qtged/src/qbrlcad/main_window.cxx): Very beginnings
of context menu for treeview items |
01:22.30 |
*** join/#brlcad yiyus
(1242712427@je.je.je) |
02:20.02 |
*** join/#brlcad kintel
(~kintel@unaffiliated/kintel) |
02:36.13 |
brlcad |
clock: that sounds like a terrible way, but
would work |
02:36.52 |
brlcad |
clock: find . - |
02:36.56 |
brlcad |
clock: find . \( -name \*.cpp -o -name \*.cxx
-o -name \*.c -o -name \*.h \) -not -regex '.*other.*' -not -regex
'.*misc.*' -not -regex '.*build.*' | xargs du -k | awk '{sum +=
$1}END{print sum}' |
02:37.20 |
brlcad |
there's a quick n' dirty tally whipped up on
the fly |
02:37.26 |
clock |
wow |
02:37.32 |
clock |
with a bonus tally! |
02:37.59 |
clock |
brlcad, so I see other misc and build
shouldn't occur in the path |
02:38.09 |
clock |
and there are cpp and cxx too |
02:38.21 |
clock |
brlcad, is BRL-CAD more in C or more in
C++? |
02:38.30 |
brlcad |
looks like we don't have any cxx any more, so
you could leave that part out |
02:38.52 |
brlcad |
well, compare the .c with the .cpp and find
out |
02:39.38 |
brlcad |
starseeker: yeah, either check for nulls or
prevent the functions from returning nulls |
02:40.39 |
brlcad |
starseeker: prior could couldn't possibly ever
be null because it was accessing a struct field ... now it's
getting passed a pointer and that pointer is being dereferenced
(looks like in just 10 or so places) without getting
checked |
02:41.25 |
brlcad |
not just a coverity warning, it'll almost
certainly end up with a crash with the right conditions (which
again just was never possible even if memory was
uninitialized) |
02:54.16 |
*** join/#brlcad fenn
(~fenn@131.252.130.248) |
02:54.17 |
*** join/#brlcad kanzure
(~kanzure@131.252.130.248) |
03:04.35 |
clock |
brlcad, I tried to build a model how the brain
understands language |
03:04.41 |
clock |
brlcad, and would like to run it on BRL-CAD
source |
03:05.04 |
clock |
because it spits out what it believes to be
the most important concepts appearing in the body of the source
code |
03:05.29 |
clock |
brlcad, I can imagine such a listing of the
topmost important things with a brief explanation or reference to
documentation could be a handy reference for BRL-CAD
programmers |
03:05.33 |
clock |
brlcad, especially beginners |
03:07.01 |
brlcad |
okay, a bit interesting to see what it comes
up with |
03:07.37 |
clock |
brlcad, I think it likes large corpora of
statistics |
03:07.40 |
clock |
like 5G overnight |
03:07.52 |
Stragus |
You want a natual language parser to run on
the BRL-CAD source? :) The documentation I guess? |
03:08.03 |
clock |
I got impression that the pure BRL-CAD source
has only 2 MB but u said according to your check something order of
20 MB |
03:08.26 |
clock |
Stragus, nono its not a natural language
parse |
03:08.46 |
brlcad |
I assume this is some sort of per file word
frequency analysis |
03:09.00 |
brlcad |
aggregated over all files |
03:09.13 |
Stragus |
The brain has pretty fixed rules on grammar
and how it understands language, that's why all human languages
share many fundamental characteristics |
03:10.29 |
Stragus |
has a friend with a master in
linguistics, quite an interesting topic |
03:11.12 |
clock |
Stragus, can you elaborate more what u in this
sentence mean under "fixed", "rules", and "grammar"? |
03:11.33 |
clock |
Stragus, and "understands"? |
03:11.46 |
clock |
Stragus, and "language"? |
03:12.00 |
clock |
Stragus, I understand your usage of the word
"brain" pretty clearly. |
03:13.07 |
Stragus |
We are born with many hardwired concepts of
grammar. Any language that wouldn't follow these rules couldn't be
processed that easily, the frontal cortex would have to do all the
painstaking parsing work |
03:13.42 |
clock |
Stragus, yes I can imagine... like feeding it
a bzipped stream might not fit the machinery of language the brain
comes with |
03:14.00 |
clock |
Stragus, so although a pretty simple program
can pase a bzip stream we might sweat pretty much doing
so |
03:14.07 |
clock |
Stragus, u mean something like that? |
03:14.10 |
Stragus |
Ah yes, the frontal cortex would have a very
hard time sorting 9000000 strings lexically :) |
03:14.21 |
Stragus |
(The BWT transform of BZIP2) |
03:14.33 |
Stragus |
But I meant the grammar |
03:15.23 |
Stragus |
Subjects and verbs, nested and relative
propositions, a lot of stuff is hardwired |
03:15.30 |
clock |
Stragus, OK so can u give a nice example of
what u said? |
03:15.45 |
clock |
Stragus, that sounds pretty
fascinating |
03:15.47 |
clock |
I mean... |
03:16.02 |
clock |
When they pumped subjects and verbs into my
head in primary school I didnt find it fascinating |
03:16.12 |
clock |
but the fact u say now there is some kind of
machinery inside for that |
03:16.33 |
clock |
my skull is pretty stiff and filled with
incompressible liquid by the way :) |
03:17.07 |
Stragus |
Linguists have created "simple" languages
based on logic (and very consistent) that don't follow our
hardwired language grammar, and they were extremely hard to learn
for anyone |
03:17.25 |
clock |
Stragus, Thai? :-P |
03:17.38 |
Stragus |
No. :) |
03:18.03 |
clock |
Stragus, yeah but do they have some kind of
schematic how those verb and subject circuits are wired? |
03:18.15 |
clock |
Stragus, like u know the entorhinal dead
reckoning decoders? |
03:18.56 |
clock |
Stragus, theres basically a hierarchy of
triangular bandlimited 2D barcodes plus some hashing |
03:19.14 |
clock |
Stragus, and it takes all kinds of data from
movement and position and turns |
03:19.41 |
clock |
And works like dead reckoning
navigation |
03:19.46 |
clock |
I mean description into that kind of
level |
03:19.55 |
Stragus |
I haven't studied the stuff, but they pretty
much found the fundamental grammatical rules any human language has
to follow |
03:20.26 |
Stragus |
Obviously it's still rather abstract,
cognitive sciences are pretty fuzzy |
03:20.34 |
clock |
I dont think they are fuzzy |
03:20.48 |
clock |
the visual front end from hubel wiesel I think
is pretty solid |
03:20.55 |
Stragus |
Well, compare psychology to physics.
:) |
03:20.58 |
clock |
although I wish they understood a it about
signal processing |
03:21.39 |
clock |
Stragus, I think psychology is a mixture of
astrology, cybernetics and programming of analog portable
supercomputers |
03:21.49 |
clock |
throw away the astrology part and you get a
nifty science in my opinion |
03:22.14 |
Stragus |
Human emotions don't have much to do with
programming and computers |
03:22.19 |
clock |
with astrology I mean Freud and similar
Science Based On Statistically Significant Repeatable Experimental
Data |
03:24.14 |
clock |
Stragus, if you have an opinion that human
emotions don't have much to do with programming and computers, u
are welcome to have your own opinion, but please dont present your
opinions as if they were facts |
03:24.31 |
clock |
Stragus, cause people often feel disrespected
when that is done |
03:24.33 |
clock |
And I am people too |
03:25.33 |
Stragus |
Software programming and computers are pure
maths, brute-force applied logic |
03:25.46 |
Stragus |
The human brain doesn't quite work like
that |
03:26.03 |
clock |
Hmm |
03:26.20 |
Stragus |
(Or any neural network, simulated ones
too) |
03:26.22 |
clock |
And everywhere I look inside brain, I see
logarithms, principal component transforms and differential
equations |
03:26.25 |
clock |
sines and cosines |
03:26.34 |
clock |
square (function) |
03:26.44 |
clock |
gaussian curves |
03:26.59 |
clock |
convolutions |
03:27.19 |
clock |
I would call that pure maths |
03:27.41 |
Stragus |
Hum... I see neurons, a lot of neurons,
connected chaotically as a learning network |
03:27.57 |
clock |
Stragus, sure thats another level |
03:28.22 |
clock |
Stragus, I think its as if I say "everywhere I
look inside Pentium I see buses, registers, adders, multipliers,
clock distributors" |
03:28.43 |
clock |
Stragus, and as if you said "Hum... I see CMOS
transistors, lots of CMOS transistors,..." |
03:29.15 |
clock |
Stragus, by the way I love those
neurons... |
03:29.30 |
clock |
One neuron can do a PCA on like 10'000 inputs
without blinking an eye |
03:29.39 |
clock |
with power consumption of like I don't know
how little |
03:29.49 |
Stragus |
Power efficiency is great, but neurons are
slow :p |
03:30.11 |
clock |
Stragus, but? |
03:30.29 |
clock |
Stragus, anyway what exactly told you that
friend about the subject and verb stuff? |
03:30.36 |
Stragus |
Great power efficiency in a dense 3D network,
with a built-in fluid for heat evacuation and nutriment transport,
but each neuron is slow |
03:31.02 |
clock |
Stragus, slow? |
03:31.16 |
Stragus |
I don't recall exactly, I'll have to ask him
to explain again what the fundamental grammar rules are |
03:31.27 |
clock |
Stragus, maybe u can show me a faster system
that can do complete transocding of objective reality into
subjective and back into objective in less than 200 ms? |
03:31.54 |
clock |
Stragus, and the total calculating power is
37'000 teraflops, with 20W power input |
03:32.18 |
clock |
Stragus, I am satisfied both with how fast my
brain is in terms of MFLOPS and in terms of latency |
03:32.37 |
clock |
Stragus, even such a "trivial" thing like a
digital video broadcasting encoder can have latencies like 5'000
ms |
03:32.47 |
Stragus |
I wonder how they would compute a processing
power in terms of billions of floating point operations per
second! |
03:33.00 |
clock |
and the fastest supercomputer NUDT Tianhe-2
has 37 MFLOPS but with power input of 5MW |
03:33.01 |
Stragus |
I mean, most humans can perhaps do 1-2 flop
per second |
03:33.26 |
clock |
Stragus, I can do more |
03:33.30 |
Stragus |
Power efficiency is amazing though, that's for
sure |
03:33.53 |
clock |
Stragus, my brain can churn 2x150Mpix 3D video
down to 2x1Mpix at about 50 Mpix-s |
03:34.00 |
clock |
with 3D decoding motion decoding |
03:34.04 |
clock |
object recognition |
03:34.08 |
clock |
and that all pretty whipping fast |
03:34.18 |
clock |
I dont have to contemplate long before I
recognize a dog, a cat, or my mother |
03:34.55 |
Stragus |
Indeed. I'm just saying that "floating point
operations per second" is a little misleading :) |
03:34.58 |
clock |
Stragus, if u can implement that in 1-2 flop
per second I would suggest you to apply your V to a company that
makes graphic cards :) |
03:35.02 |
clock |
V -> CV :) |
03:35.09 |
Stragus |
The brain's visual processing is
fantastic |
03:36.35 |
brlcad |
clock: what's your website url that has your
.g models? |
03:36.43 |
*** join/#brlcad LordOfBikes_
(~armin@dslb-092-074-238-118.092.074.pools.vodafone-ip.de) |
03:36.47 |
brlcad |
the networking device one |
03:37.09 |
clock |
brlcad, http://ronja.twibright.com/3d/ |
03:37.17 |
clock |
brlcad, for each model there is a .g clickable
link |
03:37.25 |
brlcad |
thanks |
03:37.33 |
clock |
Stragus, I can second that |
03:37.50 |
clock |
Stragus, I built it myself in software and I
find the results fantastic |
03:38.15 |
clock |
Stragus, it can do like a half of
Photoshop/GIMP just by tuning the parameters of the
decoder |
03:38.22 |
clock |
Stragus, like u have music equalizer you
know? |
03:38.28 |
clock |
Wanna sharper? Turn that knob |
03:38.37 |
clock |
wanna white balance? Throw that channel
out |
03:38.43 |
clock |
Wanna B/W? throw that one out |
03:38.56 |
clock |
brightness? contrast? edge detect? |
03:39.48 |
clock |
Oh I have a listing here... hope is not too
many lines |
03:39.49 |
Stragus |
There are so many aspects to the brain's
visual processing most of us aren't even aware of |
03:40.00 |
clock |
Colors: Levels: Gamma |
03:40.00 |
clock |
Colors: Color Balance |
03:40.00 |
clock |
Colors: Hue-Saturation |
03:40.00 |
clock |
Colors: Brightness-Contrast |
03:40.00 |
clock |
Colors: Invert |
03:40.01 |
clock |
Colors: Auto: White Balance |
03:40.03 |
clock |
Colors: Auto: Color Enhance |
03:40.05 |
clock |
Colors: Auto: Normalize |
03:40.09 |
clock |
Colors: Auto: Stretch Contrast |
03:40.11 |
clock |
Colors: Auto: Stretch HSV |
03:40.13 |
clock |
Colors: Retinex |
03:40.15 |
clock |
Filter: Blur: Gaussian Blur |
03:40.17 |
clock |
Filter: Blur: Motion Blur |
03:40.19 |
clock |
Filter: Enhance: Sharpen |
03:40.21 |
clock |
Filter: Enhance: Unsharp Mask |
03:40.22 |
Stragus |
We have two powerful motion detection filters
near the focus point, but just a lesser motion detection filter in
the peripheral vision |
03:40.23 |
clock |
Filter: Edge-Detect: Difference of
Gaussians |
03:40.25 |
clock |
Generic: Convolution Matrix (in limited
sense) |
03:40.27 |
clock |
COlor balance of skin without influencing the
white balance |
03:40.37 |
Stragus |
There are optical illusions that exploit that
stuff |
03:40.38 |
clock |
Changes of face expression, apparent age and
facial attractivenes |
03:40.49 |
clock |
Changes in emotional atmosphere of the
scene |
03:41.55 |
clock |
Stragus, in a picture with a field of plants,
it can even change the plants to a different type |
03:42.17 |
clock |
plants=vegetables, flowers etc. |
03:43.30 |
clock |
Stragus, can produce those rainbow-like LSD
colours |
03:44.47 |
Stragus |
The brain's visual processing is something we
have a very hard time matching, not by lack of processing power
but... somehow, proper code |
03:45.56 |
Stragus |
A neural network is perfectly suited to that
kind of problem though. No comparison with floating point
operations :) |
03:46.57 |
clock |
Stragus, I didn't have hard time matching
that |
03:47.01 |
clock |
Stragus, so please speak for
yourself |
03:47.27 |
clock |
Stragus, In my opinion its the sheer
processing power thats difficult to match with semiconductor
computers |
03:48.12 |
clock |
Stragus, and again, please dont present your
opinions as facts. In my model the correspondence with floating
operations was pretty almost 1:1 straightforward |
03:48.42 |
clock |
Stragus, by the way, do you know what it looks
like, if you inject random noise into the primary visual cortex
(V1)? |
03:48.57 |
Stragus |
I don't think it's just an opinion that we
don't have software matching a brain's visual processing |
03:49.13 |
Stragus |
If you didn't have a hard time matching that,
then you should share it with the world! |
03:50.11 |
clock |
Stragus, please guess what random noise into
the primary visual cortex (V1) looks like :) |
03:50.25 |
Stragus |
You mean noise in the visual input? The brain
probably filters out most of it |
03:50.46 |
clock |
not visual input, I said primary visual
cortex (V1) |
03:50.54 |
clock |
Thats back in the head |
03:51.00 |
clock |
Broadmann area 17 if I remember
correctly |
03:51.11 |
Stragus |
How do you insert noise in a neural network?
Make some neurons fire randomly? |
03:51.41 |
Stragus |
That would probably be either hallucinations
or a seizure |
03:51.47 |
clock |
Stragus, so you want to know what kind of
noise I am injecting? |
03:52.00 |
clock |
Stragus, I take all the analog channels that
are there and add additive white gaussian noise |
03:52.27 |
clock |
which has same signal to noise ratio in all of
the channels |
03:52.44 |
clock |
hehe :) |
03:52.52 |
clock |
that neither hallucination nor
seizures |
03:52.58 |
clock |
it looks like a diamond snowstorm :) |
03:53.23 |
clock |
which probably I should explain what a diamond
snowstorm is |
03:53.49 |
Stragus |
I'm not sure what analog channels you are
talking about. The neurons firing potentials? |
03:53.50 |
clock |
Stragus, have u ever been in a
snowstorm? |
03:54.08 |
Stragus |
I live in Montreal, roughly 40 times per year
:) |
03:54.33 |
clock |
Stragus, OK I define analog channel as the
value when you take the neuron axon and apply a gaussian time
window convolution with standard deviation of 20
millisecond |
03:54.51 |
Stragus |
Okay |
03:55.02 |
clock |
Stragus, have you noticed the snowstorm noise
in the picture kinda differs from TV noise= |
03:55.17 |
clock |
In the sense that there are also "bigger" or
"lower frequency" "pockets" floating around? |
03:55.38 |
Stragus |
Hum, I guess so |
03:55.52 |
clock |
and now add a colourful, rainbowy-spectrum
twinkling to that |
03:55.57 |
clock |
like stars twinkle in diamond colors |
03:56.17 |
Notify |
03BRL-CAD Wiki:Maths22 * 7765 /wiki/Deuces:
Removed completed code tasks. |
03:56.36 |
clock |
Stragus, if you apply that kind of noise over
a persons face the identity of the person appears kind
auncertain |
03:56.49 |
clock |
of course stronger noise means more
uncertainty |
03:56.59 |
Stragus |
Of course. So that's noise on the visual input
then |
03:57.07 |
clock |
Stragus, to me it looks like I recall the
persons face from memory but I cannot recall exactly |
03:57.26 |
clock |
Stragus, no, as I said, not visual input, in
V1 |
03:57.40 |
clock |
If u put noise on visual input you get
something like TV noise or digital camera noise |
03:57.47 |
clock |
That doesnt make human face look
uncertain |
03:57.52 |
clock |
that face just looks noise |
03:57.53 |
clock |
noisy |
03:57.58 |
clock |
like low quality pic |
03:58.05 |
Stragus |
At some point, with enough noise, it becomes
uncertain :) |
03:58.19 |
clock |
sure |
03:58.21 |
Stragus |
I have done some image processing, at some
point recognition becomes a probability with noise |
03:58.24 |
clock |
but this uncertainty is more natural |
03:58.30 |
clock |
like incomplete recall from memory |
03:58.40 |
clock |
or... looks more natural to me |
04:00.35 |
clock |
Stragus, I get pretty disgusted by so called
image processing |
04:00.44 |
clock |
Stragus, for example what they call "edge
detectors" |
04:00.53 |
clock |
Seems to me people have never seen an edge in
their life |
04:00.59 |
clock |
or have complete lack of idea what an edge
is |
04:01.04 |
Stragus |
Ah? :) I have done some of that |
04:01.18 |
clock |
and those filters they apply and claim to be
edge detectors.. I think they are everything but edge
detectors |
04:01.25 |
clock |
and when then I see those broken, noisy
results |
04:02.03 |
Stragus |
Edge detection is just a little half-broken
step on the way to the next processing steps, it doesn't have to be
accurate |
04:02.28 |
clock |
Stragus, OK :) |
04:02.42 |
clock |
Please design me a sparrow with your "edge
detection that doesnt have to be accurate" |
04:02.49 |
clock |
In the size of a sparrows head |
04:02.54 |
clock |
with the power available to sparrows
brain |
04:03.06 |
Stragus |
Eheh |
04:03.27 |
clock |
youre supposed to fly at 30km/h through a 5cm
channel between random, ambiguous looking twigs of a bush, with
such a reliability you dont kill yourself more often than once per
20 years |
04:03.32 |
Stragus |
Neural networks don't do any of that because
it's an entirely different way of processing information |
04:03.43 |
Stragus |
But edge detection works okay for transistor
chips |
04:04.54 |
clock |
Stragus, In my opinion neural networks do
that |
04:05.14 |
Stragus |
Edge detection? |
04:05.18 |
clock |
yes |
04:06.10 |
clock |
The trick is, what brain calls an edge is
something quite different than what "image processing" people call
edge |
04:06.17 |
Stragus |
Well not the computer science definition of
edge detection, but at some point it tracks the edges between
different objects, of course |
04:06.57 |
Notify |
03BRL-CAD:starseeker * 63364
brlcad/branches/qtged/src/qbrlcad/cadtreeview.cxx: Try a slightly
different approach to forcing a full update of related object
items |
04:07.13 |
clock |
Stragus, if you want to know what brain calls
an edge |
04:07.23 |
clock |
I recommend look around you and find some
edges |
04:07.32 |
clock |
those are what brain calls edges :) |
04:07.34 |
clock |
So simple :) |
04:07.41 |
clock |
You may contemplate what they have in
common |
04:08.01 |
clock |
Stragus, and please dont try to limit your
choice of edges to what you know an "image processing" algorithm
would have easy time dealing with |
04:08.14 |
Stragus |
That doesn't explain how the brain processes
the information :), the visual cortex is autonomous and we only
consciously get the final results |
04:08.26 |
Stragus |
No way to debug the process and examine the
registers and stack memory |
04:08.54 |
clock |
Stragus, no? |
04:09.32 |
clock |
Stragus, I believe one holiday postcard has
the whole structure of human visual cortex encoded in (except the
motion part of course) |
04:09.39 |
Stragus |
As far as I know, we don't yet understand the
brain's visual processing too well |
04:09.45 |
clock |
Stragus, to figure that, all you have to do is
look at the postcard |
04:09.57 |
clock |
the postcard is not a white noise of
statistically independent random pixels |
04:11.20 |
Stragus |
We have identified the zones that fire up when
doing some specific visual tasks |
04:11.52 |
clock |
which is kinda like identifying the zones in
PC which fire up when you play games or back your hard
disk |
04:11.52 |
Stragus |
But I don't think we have understood how the
image processing works exactly |
04:12.00 |
Stragus |
Yup |
04:12.43 |
clock |
Stragus, I recommend, just try to build it
:) |
04:12.55 |
clock |
Wanna understand motorcycles? Build a
motorcycle :) |
04:13.07 |
Stragus |
So I should build a visual cortex?
:) |
04:13.09 |
clock |
Wanna understand brain? Build a brain
:) |
04:13.12 |
clock |
yep |
04:13.36 |
clock |
Stragus, thats exactly what I did |
04:13.50 |
clock |
I was curious and wanted to see what happens
when I manipulate the signals in primary visual cortex |
04:13.58 |
Stragus |
I had a hard time writing code that would
locate photogrammetry targets in an image, through shadows, noise,
fuzzy, out of focus and so on :p |
04:14.10 |
clock |
So I googled up all kinds of scientific
studies on that topic and then programmed the things they said the
brain does with the image data |
04:14.13 |
Stragus |
"Hard time" might be an exageration but it
wasn't exactly easy |
04:14.24 |
clock |
And in that process I felt like I understand
the way the brain is designed |
04:14.50 |
Stragus |
Cool. I haven't digged enough on the
topic |
04:15.12 |
clock |
The problem the studies are done by people who
havent much clue about the topic |
04:15.13 |
clock |
biologists |
04:15.28 |
clock |
they write a minutiously precise treatise on
how the brain reacts on white lines |
04:15.31 |
Stragus |
There are so many interesting topics to learn,
so many interesting projects to explore, and so little time. And a
daily job. |
04:15.38 |
clock |
and forget to tell you how it reacts to grey
lines |
04:15.54 |
clock |
its like... |
04:16.10 |
clock |
if Top Gear did a car review where all they do
is drive straight full gas |
04:18.15 |
clock |
Stragus, I must say, the implementation of
image processing in the brain is VIP ***** 5 star Grand
Deluxe |
04:18.24 |
clock |
Like the best filters I can imagine |
04:18.27 |
clock |
the best algorithm |
04:18.31 |
clock |
I could hardly imagine it better |
04:18.36 |
clock |
sometimes it was kinda breathtaking |
04:18.38 |
clock |
and the elegance |
04:18.58 |
clock |
It was like |
04:19.27 |
clock |
if you have 80 billion dollars and want to
flaunt your wealth by having a image processing system designed for
you |
04:19.32 |
clock |
then I would do exactly that |
04:19.43 |
Stragus |
Yup, a couple millions year of evolution,
computer science hasn't caught up yet :) |
04:19.43 |
clock |
louis vuitton of image processing |
04:19.43 |
clock |
diamond clad |
04:19.43 |
clock |
gold plated |
04:19.58 |
clock |
individually signed |
04:20.29 |
Notify |
03BRL-CAD Wiki:Maths22 * 7766 /wiki/Deuces:
Removed completed documentation items. |
04:20.36 |
clock |
Stragus, have you noticed your vision is
neither pixelated, not stuttering, blurring on movement, no stairs,
squary artifacts, nothing seen in computer video? |
04:20.40 |
Stragus |
Really? Generally in biology, some stuff are
strangely designed, as a by-product of an evolutionary process
blind to long-term goals |
04:21.21 |
Stragus |
Yes well, the eye isn't made of pixels
either |
04:21.36 |
clock |
in my opinion it is |
04:21.46 |
clock |
the individual photocells |
04:21.46 |
Stragus |
And the visual cortex smoothes that input to
extract the real information |
04:21.52 |
Stragus |
And that's what we really "see" |
04:22.17 |
Notify |
03BRL-CAD Wiki:Maths22 * 7767 /wiki/Deuces:
Removed completed QA task |
04:22.25 |
clock |
it is because there is a massive
downsampling |
04:22.40 |
clock |
to a matter the signal to noise ratio to the
antialising noise is HiFi |
04:22.55 |
clock |
antialiasing -> aliasing |
04:23.03 |
clock |
and it has no time quantization |
04:23.09 |
clock |
no issues with frame rate |
04:23.17 |
Stragus |
There are no frames :) |
04:23.28 |
Stragus |
It's continuously smooth processing |
04:23.37 |
clock |
in time domain yes |
04:23.46 |
clock |
in spatial no but it has such a brutal
oversampling its like... |
04:23.55 |
Stragus |
Yup |
04:24.13 |
clock |
if the boss told you instead try to design it
as cheap as possible as expensive as possible |
04:24.38 |
Stragus |
thinks we need a switch to
turn off the downsampling and read the raw input, when we need to
read that street name 200 meters away |
04:25.45 |
clock |
Stragus, no no its not like that |
04:26.06 |
clock |
I said wrong |
04:26.11 |
clock |
its not downsampling its
oversampling |
04:26.19 |
clock |
it doesnt cause any loss of
resolution |
04:26.55 |
clock |
the retina is 150 Mpix |
04:27.02 |
clock |
the lens is 10mm not a great
sharpness |
04:27.11 |
clock |
something like put a 150Mpix CMOS into your
$30 webcam |
04:27.45 |
clock |
so you have plenty extra resolution on the
sensor to CUT THE CRAP WITH SPATIAL SAMPLING ONCE AND
FOREVER! |
04:28.04 |
clock |
who cares it takes huge amounts of extra
processing power? |
04:28.13 |
clock |
We have 37'000 Teraflops onboard :) |
04:29.00 |
Notify |
03BRL-CAD Wiki:Maths22 * 7768 /wiki/Deuces:
Removed completed UI tasks |
04:33.20 |
Stragus |
I have a very hard time measuring the brain as
"floating point operations per second" :p |
04:33.57 |
clock |
Stragus, here is an example from my simulator
what happens when you turn up the colors extremely http://i.iinfo.cz/images/157/leaves.jpg |
04:34.20 |
clock |
I find it facsinating |
04:34.34 |
clock |
unlike similar effect in Photoshop/GIMP it
doesn't appear artificial to me at all |
04:34.37 |
clock |
very naturak |
04:34.39 |
clock |
natural |
04:34.44 |
clock |
not in any way "clipped" |
04:35.03 |
Notify |
03BRL-CAD Wiki:Maths22 * 7769 /wiki/Deuces:
Removed completed outreach/research tasks. |
04:35.11 |
clock |
I kinda don't even have a feeling it should
show something different than the original scene |
04:35.21 |
clock |
let apart the fact its kinda psychedeleic or
halucinogenic |
04:35.50 |
Stragus |
I find the second image more difficult to
recognize. Like the leaf on top of the other leaf creating a
shadow |
04:36.25 |
clock |
Stragus, OK I can understand that |
04:36.39 |
clock |
Stragus, for me the essence or atmosphere of
the summer is greatly amplified in the second image :) |
04:37.32 |
clock |
its like the red heat of the sun is almost
unbearable on the leaves |
04:37.41 |
clock |
and the coolness of the shadows in the
backround extreme :) |
04:37.52 |
Stragus |
Eh. The image processing I have done was stuff
like that : http://www.rayforce.net/photogrammetry006.png |
04:38.00 |
clock |
for me its kinda ... summer to the maxxx!
:) |
04:38.06 |
Stragus |
So... basically locate and identify
photogrammetry targets in the image |
04:39.09 |
clock |
Stragus, congrats, looks very cleanresults!
:) |
04:39.22 |
clock |
Stragus, why did u throw out the
colors? |
04:39.34 |
clock |
Stragus, the brain doesnt have colours in
picture just for fun :) |
04:39.35 |
Stragus |
Eh yes, it works through shadows, blurriness,
fuzziness, and all kind of other problems |
04:40.00 |
Stragus |
I didn't throw out the colors, it's just
easier to see the debugging output over grayscale |
04:40.12 |
Stragus |
The code works with real RGB |
04:40.16 |
clock |
ah good :) |
04:41.36 |
clock |
Stragus, my brain identifies a photogrammetry
target upper left from #21 and #9 |
04:41.47 |
Stragus |
I know that! :) |
04:41.56 |
clock |
Stragus, is my brain wrong or your code
wrong? |
04:42.03 |
Stragus |
But since it's partly occluded, it can't be
sure of the bit decoding around the core |
04:42.37 |
clock |
Stragus, my brain can locate the core pretty
accurately :) |
04:42.38 |
Stragus |
A smarter code would be able to figure out
what really belongs to the coding ring and what not |
04:42.51 |
clock |
Stragus, and you even haven't defined to me
what a photogrammetry target should look like? |
04:42.53 |
Stragus |
Yes yes, it's the ring that can be
tricky |
04:43.12 |
clock |
oh another between 10 and 17 |
04:43.47 |
clock |
but OK I cheat with 37 PFLOPS :) |
04:44.21 |
Stragus |
Can you identify all targets in 50 such images
per second? :) |
04:44.29 |
Stragus |
Well, "almost all targets" |
04:44.34 |
Stragus |
Non-occluded ones anyway :p |
04:44.35 |
clock |
no :) |
04:45.41 |
*** join/#brlcad gaganjyot
(~gaganjyot@124.253.225.90) |
04:45.53 |
clock |
Stragus, but I can drink from a glass without
spilling it |
04:46.01 |
Stragus |
My code can't do that |
04:46.02 |
clock |
ride a skateboard with less than 1 crash per
year |
04:46.51 |
clock |
and chat on IRC while having an intention to
actually do gym instead :) |
04:47.33 |
Stragus |
"A human being should be able to change a
diaper, plan an invasion, butcher a hog, conn a ship, design a
building, write a |
04:47.36 |
Stragus |
sonnet, balance accounts, build a wall, set a
bone, comfort the dying, take orders, give orders, cooperate, act
alone, |
04:47.40 |
Stragus |
solve equations, analyze a new problem, pitch
manure, program a computer, cook a tasty meal, fight efficiently,
die gallantly." |
04:47.46 |
Stragus |
Specialization is for computers. |
04:49.44 |
clock |
Stragus, can your program find a
boyfriend? |
04:49.46 |
clock |
Stragus, I can't do that |
04:50.52 |
Stragus |
I heard some people have created algorithms
for that... :) |
04:51.31 |
clock |
Stragus, my Thai ex boyfriend was
amazing |
04:51.48 |
clock |
he could do 37 PFLOPS with just 66% of power
consumption of me |
04:51.54 |
clock |
cause he was only 50kg heavy |
04:52.01 |
clock |
and the power consumption depends on weight
:) |
04:52.26 |
clock |
He called himself stupid because he had only
primary school |
04:52.32 |
clock |
But I call this concentrated
intelligence |
04:52.41 |
Stragus |
There are many ways to learn |
04:53.00 |
Stragus |
I only have high-school myself, I just learned
the rest on my own |
04:54.35 |
clock |
Stragus, u know what human life is
about? |
04:54.41 |
clock |
I think I can tell you that pretty exactly
:) |
04:54.52 |
clock |
searches in his
reference |
04:57.16 |
clock |
its about things, rooms, home, people, cars,
work, houses, men, time, places, areas, water, women, class,
shows,... |
04:58.27 |
Stragus |
Darn. I got it all wrong |
04:58.48 |
Stragus |
I thought it was about creating fun stuff,
learning fun stuff, working a little on the side, and eating good
food |
05:00.20 |
clock |
Stragus, maybe your life is statistically
unusual :) |
05:00.28 |
clock |
Stragus, but I can confirm my life too is
about those things :) |
05:00.55 |
clock |
and things is #1 |
05:01.05 |
clock |
I think no other species is so obsessed with
things as homo sapiens |
05:01.20 |
clock |
like the amount of things I have at
home... |
05:01.29 |
clock |
compared to how many things a crow or a dog
have... |
05:01.29 |
Stragus |
Most other species are obsessed with food. But
that's less a problem for us, so we found other stuff to worry
about |
05:01.33 |
clock |
probably 0 :) |
05:01.37 |
clock |
a collar maybe and a bowl :) |
05:02.32 |
clock |
a kennel :) |
05:06.12 |
clock |
Stragus, the list is by the way output of my
language analyzer modeled by the brain |
05:06.21 |
clock |
on dream database |
05:06.32 |
clock |
if u run it on project gutenberg english prose
you get very similar list |
05:06.44 |
clock |
I actually threw away personal pronouns which
lead the list |
05:06.48 |
maths22 |
brlcad: I'm back, and my screen session was
still live! |
05:06.56 |
clock |
and no, it is not word frequency, as brlcad
suggested |
05:07.03 |
clock |
if it were, the leader would be probably
"the" |
05:07.12 |
clock |
But "the" doesnt figure in the top words at
all |
05:07.58 |
clock |
it works prefectly same well on Khmer language
which doesn't have any spaces |
05:08.17 |
clock |
and therefore you cannot know the word
boundaries and they are not fed explicitly :) |
05:19.06 |
gaganjyot |
brlcad, sincere apologies for not doing it
yesterday |
05:19.13 |
gaganjyot |
https://github.com/gaganjyot/LCKernel-Example |
05:19.25 |
gaganjyot |
here is the clone of sketcher file in
proc-db |
05:20.01 |
gaganjyot |
it was very easy. I was only unable to
understand regarding the export to G file which we discussed
yesterday hence made it late :( |
05:21.47 |
clock |
brlcad, why are you asking about the .g files
from Ronja? :) |
05:27.11 |
clock |
brlcad, now I confirm I get 26 MB with your
method of selecting source |
05:27.19 |
clock |
brlcad, I must have done something
wrong |
05:33.13 |
brlcad |
clock: I was just looking for the site but
couldn't remember the name |
05:35.24 |
brlcad |
gaganjyot: heh, any particular reason you put
that all into a class with just one big "do it" function instead of
just embedding the logic in main so they could be compared
side-by-side? :) |
05:38.43 |
brlcad |
gaganjyot: either way, this looks pretty good.
someone willl need to de-c++11ify it in order to really see how
this will map into our backend lib, and remove the namespace using
declarations |
05:42.59 |
*** join/#brlcad gaganjyot
(~gaganjyot@124.253.225.90) |
05:44.52 |
clock |
brlcad, Stragus OK so what my program
say |
05:45.27 |
clock |
<PROTECTED> |
05:46.02 |
clock |
hit dist dir ptr size struct min bu_log( edge
region curve new_ mat id data tmp num loop scale _name nmg_ vertex
path result faces fastf_t |
05:46.32 |
clock |
width plane height temp =0; rt_
[1] = ray it pixel pp _list left model solid ... |
05:47.01 |
clock |
also: -2013 United States Government as
represented by the U.S. Army Research Laboratory... :) |
05:47.30 |
clock |
brlcad, you cannot hide before the prying eyes
of my program that BRL-CAD has something to do with "United States
Government as represented by the U.S. Army Research Laboratory"
:) |
05:47.32 |
clock |
hahahaha :) |
05:48.23 |
clock |
brlcad, shouldnt one update the -2013 to
-2014? |
05:50.00 |
clock |
and cryptic "f, -0.3" |
05:50.11 |
clock |
I wonder what that number has in relation with
BRL-CAD :) |
05:50.29 |
clock |
bu_vls_addr(& :) |
05:50.32 |
Notify |
03BRL-CAD Wiki:Inderpreet * 7770 /wiki/Deuces:
/* Code */ |
05:51.00 |
clock |
bu_free(, segment, parent, YERO( |
05:51.02 |
clock |
ZERO |
05:51.35 |
Notify |
03BRL-CAD Wiki:Inderpreet * 7771 /wiki/Deuces:
/* Code */ added new task for GCI |
05:52.11 |
clock |
BU_ALLOC(, VSCALE, BOT... |
05:52.34 |
clock |
VSUB2, faceuse, corner, ellipse |
05:53.01 |
clock |
poly, command, eu2, sketch |
05:53.39 |
brlcad |
it is -2014 |
05:55.49 |
clock |
torus failure |
05:55.51 |
clock |
LOOOOOOOOL |
05:56.02 |
clock |
"While driving a car, I suffered a torus
failure" :)))))))) |
05:57.15 |
clock |
brlcad, I took the most recent I found for
download, brlcad-7.24.2, so maybe that version was prepared in
-2013? |
05:57.26 |
brlcad |
nope |
05:59.25 |
clock |
brlcad, everywhere I Look into the code I see
-2013 :) |
05:59.27 |
brlcad |
in fact, there's only 23 files I see that have
that string |
06:00.24 |
brlcad |
ahh, I see .. there's a disconnect |
06:00.47 |
clock |
disconnect? |
06:01.27 |
brlcad |
yeah, the version is updated on trunk, but for
some branch-merge reason, that commit isn't on the release/stable
branches |
06:01.54 |
brlcad |
s/version/date/ |
06:02.46 |
brlcad |
ahhh, I know why that is |
06:03.03 |
brlcad |
Yeah.... 7.24.2 only has commits through
Dec2013 merged into it |
06:03.12 |
brlcad |
the date update happened on new years
day |
06:03.31 |
brlcad |
that's kinda funny |
06:04.20 |
clock |
(struct ged *gedp |
06:04.43 |
clock |
looks like passing struct ged *gedp as a first
parameter is kinda popular in BRL-CAD :) |
06:05.18 |
clock |
or we also have BU_ALLOC. What can it do? I
guess allocate something :) |
06:06.03 |
clock |
BU_LIST_INIT, NEAR_ZERO( |
06:06.28 |
clock |
struct rt_ |
06:06.45 |
clock |
I guess there must be a non-negligible usage
of structures whose names begin with rt_ :) hehehe :) |
06:07.26 |
clock |
bu_vls_free, quat. I wonder if quat are
quaternions :) |
06:07.27 |
gaganjyot |
brlcad, can I participate as mentor in GCI
this year ? |
06:07.46 |
clock |
_specific, theta, bu_bomb( |
06:09.54 |
gaganjyot |
and you may check the lckernel code at github
link I sent you https://github.com/gaganjyot/LCKernel-Example |
06:13.17 |
clock |
is satisfied with the result
of the analysis |
06:45.21 |
brlcad |
gaganjyot: the criteria was spelled out in the
e-mail :) |
07:00.59 |
*** join/#brlcad gaganjyot
(~gaganjyot@124.253.225.90) |
08:00.34 |
Ch3ck_ |
brlcad: would like to know if unit tests could
be valid GCI tasks? |
08:47.28 |
*** join/#brlcad ries
(~ries@D979EA84.cm-3-2d.dynamic.ziggo.nl) |
11:37.12 |
*** join/#brlcad ries
(~ries@D979EA84.cm-3-2d.dynamic.ziggo.nl) |
11:48.58 |
*** join/#brlcad gaganjyot
(~gaganjyot@124.253.225.90) |
12:19.30 |
*** join/#brlcad clock
(~clock@212.203.58.127) |
12:48.22 |
*** join/#brlcad gaganjyot
(~gaganjyot@124.253.225.90) |
12:48.46 |
*** join/#brlcad kintel
(~kintel@unaffiliated/kintel) |
12:52.09 |
*** part/#brlcad gaganjyot
(~gaganjyot@124.253.225.90) |
13:12.42 |
*** join/#brlcad kintel
(~kintel@unaffiliated/kintel) |
13:29.02 |
*** join/#brlcad kanzure
(~kanzure@131.252.130.248) |
13:29.23 |
*** join/#brlcad fenn
(~fenn@131.252.130.248) |
14:54.39 |
*** join/#brlcad teepee
(~teepee@gateway/tor-sasl/teepee) |
15:06.05 |
Notify |
03BRL-CAD:starseeker * 63365
(brlcad/branches/qtged/src/qbrlcad/cadtreemodel.h
brlcad/branches/qtged/src/qbrlcad/cadtreeview.cxx
brlcad/branches/qtged/src/qbrlcad/cadtreeview.h): Add the ability
to the treeview to feed a path in as a string for expansion and
selection. |
15:10.48 |
*** join/#brlcad ries
(~ries@D979EA84.cm-3-2d.dynamic.ziggo.nl) |
16:11.16 |
*** join/#brlcad LordOfBikes
(~armin@dslb-092-074-238-118.092.074.pools.vodafone-ip.de) |
16:22.48 |
*** join/#brlcad ries
(~ries@D979EA84.cm-3-2d.dynamic.ziggo.nl) |
16:33.12 |
*** join/#brlcad gaganjyot
(~gaganjyot@124.253.225.90) |
16:57.32 |
Notify |
03BRL-CAD:Gauravjeet * 0
/wiki/User:Gauravjee: |
20:19.32 |
*** join/#brlcad kintel
(~kintel@unaffiliated/kintel) |
20:36.34 |
*** join/#brlcad kintel
(~kintel@unaffiliated/kintel) |
22:41.29 |
*** join/#brlcad kintel
(~kintel@unaffiliated/kintel) |
23:15.03 |
*** join/#brlcad teepee
(~teepee@gateway/tor-sasl/teepee) |
23:17.30 |
brlcad |
Ch3ck_: yes they can, but they have to be
somewhat specifically scoped (what specific function(s)) |
23:17.51 |
brlcad |
in the past, unit tests have been problematic
(bad tests, wrong tests, too hard, too much, etc) |
23:20.25 |
brlcad |
better ones have been where we provided a
specific relevant example or template, and we can define exactly
what they need to test |
23:26.00 |
kanzure |
brlcad: i was considering offering mentoring
time, but then i rembered how poorly i executed on that last time.
how desperate is the lack of mentors this time around? |