1.1 --- a/.hgignore	Sun Oct 13 12:06:09 2013 -0400
     1.2 +++ b/.hgignore	Sun Oct 13 12:09:05 2013 -0400
     1.3 @@ -1,3 +1,4 @@
     1.4  syntax: glob
     1.5  src*
     1.6  html*
     1.7 +.git/*
     1.8 \ No newline at end of file

     2.1 Binary file images/adelson-checkerboard.jpg has changed

     3.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     3.2 +++ b/org/adelson-notes.org	Sun Oct 13 12:09:05 2013 -0400
     3.3 @@ -0,0 +1,98 @@
     3.4 +#+title: Notes for "Special Topics in Computer Vision"
     3.5 +#+author: Robert McIntyre
     3.6 +#+email: rlm@mit.edu
     3.7 +#+description: 
     3.8 +#+keywords: 
     3.9 +#+SETUPFILE: ../../aurellem/org/setup.org
    3.10 +#+INCLUDE: ../../aurellem/org/level-0.org
    3.11 +#+babel: :mkdirp yes :noweb yes :exports both
    3.12 +
    3.13 +* Fri Sep 27 2013
    3.14 +
    3.15 +  Lambertian surfaces are a special type of Matt surface. They reflect
    3.16 +  light in all directions equally. They have only one parameter, the
    3.17 +  amount of energy that is absorbed/re-emitted.
    3.18 +
    3.19 +  [[../images/adelson-checkerboard.jpg]]
    3.20 +  #+caption: Lol checkerboard illusion.
    3.21 +
    3.22 +  Look into Helmholtz' stuff, it might be interesting. It was the
    3.23 +  foundation of both vision and audition research. Seems to have took
    3.24 +  a sort of Baysean approach to inferring how vision/audition works.
    3.25 +  
    3.26 +  - Homomorphic filtering :: Oppenhiem, Schafer, Stockham, 1968.  also
    3.27 +       look at Stockham, 1972.
    3.28 +
    3.29 +  Edwin Land was Adelson's hero back in the day. He needed to create a
    3.30 +  color photo for the Polaroid camera. In order to process for
    3.31 +  automatic development of film, he had to get a good approximation for
    3.32 +  the illumination/reflectance decomposition that humans do, which he
    3.33 +  called Retinex.
    3.34 +
    3.35 +  Cornsweet square wave grating is cool.
    3.36 +
    3.37 +  - Retinex :: use derivatives to find illumination. Sort of
    3.38 +               implicitly deals with edges, etc. Can't deal with
    3.39 +               non-lambertian objects. 
    3.40 +
    3.41 +
    3.42 +  Adelson introduces the problem as an "inverse" problem, where you
    3.43 +  try to "undo" the 3-d projection of the world on your retina.
    3.44 +
    3.45 +  On the functional view of vision : "What it takes" is to build a
    3.46 +  model of the world in your head. The bare minimum to get success in
    3.47 +  life is to have a model of the world. Even at the level of a single
    3.48 +  cell, I think you still benefit from models.
    3.49 +
    3.50 +  Spatial propagation is ABSOLUTELY required to separate embossed
    3.51 +  stuff from "painted" stuff. Edges, likewise, MUST have spatial
    3.52 +  context to disambiguate. The filters we use to deal with edges must
    3.53 +  have larger spatial context to work, and the spatial extent of this
    3.54 +  context must be the ENTIRE visual field in some cases!
    3.55 +
    3.56 +------------------------------------------------------------
    3.57 +
    3.58 +** Illumination, shape, reflectance all at once
    3.59 +   
    3.60 +   What if we tried to infer everything together? Some images are so
    3.61 +   ambiguous it requires propagation from all three qualities to
    3.62 +   resolve the ambiguity.
    3.63 +
    3.64 +   Brain has a competing painter, sculptor, and gaffer which each try
    3.65 +   to "build" the things in the world. There is a cost to everything
    3.66 +   such as paints, lights, and material, and then you try to optmize
    3.67 +   some cost function using these primitives.
    3.68 +
    3.69 +
    3.70 +   Horn, technical report, 1970
    3.71 +
    3.72 +
    3.73 +
    3.74 +* Fri Oct  4 2013
    3.75 +
    3.76 +  Student report. Talked about how you capture the appearance of a
    3.77 +  grape. It's actually quite compicated, involving gloss, spatial
    3.78 +  context, etc.
    3.79 +
    3.80 +  Turbosquid seems interesting. They sell 3D models of stuff.
    3.81 +
    3.82 +  BRDF -- bi-directional reflectance distribution function this shows
    3.83 +  how a surface will behave given lighting conditions. Lambertian is a
    3.84 +  simple parameterized instantiation of this.
    3.85 +
    3.86 +  BSSRDF -- (SS = subsurface) 3D analogue of BRDF
    3.87 +
    3.88 +  What would the 3D analiogue of texture be?
    3.89 +
    3.90 +  (a : b : c) as (a + b + c : b + c : c) <-- this is just the golden
    3.91 +  ratio again!
    3.92 +
    3.93 +  CURET BTF Database lol what's this
    3.94 +
    3.95 +  This student went and gathered 1000 images of different large
    3.96 +  objects made of different materials. The images were gathered off of
    3.97 +  Flikr. 
    3.98 +
    3.99 +  
   3.100 +
   3.101 +  

     4.1 --- a/org/comprehend-singularity.org	Sun Oct 13 12:06:09 2013 -0400
     4.2 +++ b/org/comprehend-singularity.org	Sun Oct 13 12:09:05 2013 -0400
     4.3 @@ -27,4 +27,4 @@
     4.4  
     4.5  So, far in the future, when synthetic life rules the stars, it is
     4.6  likely that their behavour will still be describable in the simplistic
     4.7 -terms of the human stories of old.
     4.8 \ No newline at end of file
     4.9 +terms of the human stories of old.

     5.1 --- a/org/ideas.org	Sun Oct 13 12:06:09 2013 -0400
     5.2 +++ b/org/ideas.org	Sun Oct 13 12:09:05 2013 -0400
     5.3 @@ -33,20 +33,44 @@
     5.4  
     5.5  - aurellem shirt :: I should make an aurellem star symbol tee-shirt. 
     5.6  
     5.7 +- high school science :: this is a lesson in scientific ethics. The
     5.8 +     goal is to calculate /g/, the local gravitational
     5.9 +     acceleration. The students are told that the textbook says it's
    5.10 +     /exactly/ 9.81 before they start the experiment. See how they
    5.11 +     doctor their results to get closer to the textbook value. It's
    5.12 +     neat because for any given school, /g/ is probably *not* exactly
    5.13 +     equal to 9.81, because that is just an average!
    5.14 +
    5.15 +- opencourseware subtitles :: there are ladies who type up lectures
    5.16 +     while they are being given. These recordings should be kept and
    5.17 +     given to OCW for subtitles. If the timestamps of keys are
    5.18 +     recorded, then it is easy to make subtitles.
    5.19 +
    5.20 +- screen locking timing :: you use your computer camera to see if you
    5.21 +     are sitting in front of the computer. If you are, then the screen
    5.22 +     will never lock. If you are, then the screen will lock with a
    5.23 +     30-40 second timeout. It's an extension of using inactivity to
    5.24 +     initiate the countdown, just with more information.
    5.25 +
    5.26 +- mirror toilet :: a toilet with a square basin made or mirror instead
    5.27 +                   or porcelain. That way, you can see how good of a
    5.28 +                   wipe job you have done / watch how your excretion
    5.29 +                   system works.
    5.30 +
    5.31  - test dummies :: why don't we clone encephalic humans and use then to
    5.32                    test /in vivo/ human organ systems and drugs? It
    5.33                    would be ethical as long as there are women who are
    5.34                    willing to host the clones, and it would be a
    5.35 -                  trememdous resource for studying the human body. I
    5.36 +                  tremendous resource for studying the human body. I
    5.37                    see nothing wrong with it morally, since no one is
    5.38 -                  suffering, and it stands to save many lives throught
    5.39 +                  suffering, and it stands to save many lives throughout
    5.40                    more advanced technology.
    5.41  
    5.42  - X-ray telepresence :: given that a doctor is operating on a patient
    5.43       via telepresence, one cool things you can do is shine X-rays into
    5.44       the patient to view the insides during real time. If the system
    5.45 -     was coupled with a baysean model of the layout of the structure,
    5.46 -     and the x-rays were only fired whenever the uncertaintity of the
    5.47 +     was coupled with a Baysean model of the layout of the structure,
    5.48 +     and the x-rays were only fired whenever the uncertainty of the
    5.49       model reached a certain threshold, then the radiation damage
    5.50       and surgery risk could be minimized.
    5.51  
    5.52 @@ -58,21 +82,21 @@
    5.53       retain their superfluidity, how would the dynamics change during
    5.54       perfusion of a superfluid, where the fluid gains and looses
    5.55       superfluidity as it goes deeper into the body and is cooled by
    5.56 -     superfluid from upstresm. In summary there are two things to
    5.57 +     superfluid from upstream. In summary there are two things to
    5.58       simulate 1.) replace all blood in human with superfluid
    5.59       instantly. 2.) perfuse superfluid into human.
    5.60  
    5.61  - projective guessing :: I think that we read and see things by
    5.62 -     makeing a really good guess about what we're expecting to see,
    5.63 +     making a really good guess about what we're expecting to see,
    5.64       and then searching for our guess in what we see. If it really
    5.65       doesn't match, then we start to make more guesses / analyze the
    5.66       image from first principles, but most stuff is projective
    5.67       guessing.
    5.68  
    5.69 -- Intestinal flora maintainence :: why not innoculate babies at birth
    5.70 +- Intestinal flora maintenance :: why not inoculate babies at birth
    5.71       with "ideal" gut flora instead of whatever bullshit they
    5.72       naturally get, thus giving them optimal digestive/nutrient
    5.73 -     extraction capabilities. Might also be abot to make their farts
    5.74 +     extraction capabilities. Might also be able to make their farts
    5.75       not stink for life, too.
    5.76  
    5.77  - server culture : mirrors :: make a distributed system where people
    5.78 @@ -80,16 +104,16 @@
    5.79       the server costs of favored websites. This could make popular
    5.80       websites run at no cost. The system would require that the
    5.81       mirrored content be the same as the official source. Sort of like
    5.82 -     bittorrent for websites.
    5.83 +     bit-torrent for websites.
    5.84  
    5.85  - map programming :: one problem with functional programming is that
    5.86       in order to remain functional, you have to pass up arguments up
    5.87 -     into each calling function to get the full range of behaviour
    5.88 +     into each calling function to get the full range of behavior
    5.89       from the lower level functions. Normally people come to a
    5.90 -     compromize involving abstraction and sparing use of dynamic
    5.91 -     variables to configure runtime behaviour. What would be the
    5.92 +     compromise involving abstraction and sparing use of dynamic
    5.93 +     variables to configure runtime behavior. What would be the
    5.94       advantages of making a programming language where every function
    5.95 -     recieves one argument, a map, which contains all the symbol
    5.96 +     receives one argument, a map, which contains all the symbol
    5.97       bindings it would ever need? This map is passed on to all
    5.98       subordinate functions. This way, you could replace functions on
    5.99       the fly, and arrange for there to be sensible defaults,
   5.100 @@ -120,7 +144,7 @@
   5.101       fixatives are highly toxic, but current vitrification procedures
   5.102       do this anyway since there can be a lot of freezing damage.
   5.103  
   5.104 -- dialated security cmaera :: a security camera that would capture
   5.105 +- dilated security camera :: a security camera that would capture
   5.106       full video footage of everything at 60fps but then decide to keep
   5.107       only every 1 frame every 5 seconds unless there's something
   5.108       "interesting" happening.