Can robots do science?
I've been recently thinking about pragmatism (see William James) and its implications for the field of robotics. In particular, I'm interested in the pragmatist treatment of truth. The down-to-earth view that reality is created from truths as opposed to the traditional (rationalistic) view that truths correspond to an external reality can be utilized to develop robust algorithms in the field of robotics. On a representational level, the philosophical doctrine that truths are deemed as 'true' only when they are 'useful' means that the type of world-state information that needs to be stored/processed/considered by a thinking machine is the information relevant to the action being considered.
I will elaborate on this in the near future...
Deep Learning, Computer Vision, and the algorithms that are shaping the future of Artificial Intelligence.
Thursday, September 20, 2007
Tuesday, August 28, 2007
made the switch to Mac OS X
I recently purchased a new laptop, namely a 15 inch Macbook Pro (2.4 Ghz with stock options). I've had this laptop less than one week and already I'm very glad I've made the switch to Mac OS X. Being a long time linux user, I feel very comfortable using this machine. Although I might load up terminal more often than my girlfriend who will use the full Adobe suite of tools on her MBP, this OS definitely caters to both of us. I love the 'feels like *nix' feeling when using OS X. It was painless to get my .emacs and .screenrc files up and running. It feels good to not be running Vista right now.
Thursday, August 16, 2007
Monday, July 16, 2007
BMVC 2007: Improving Spatial Support for Objects via Multiple Segmentations
Here are some images generated from my BMVC 2007 paper.
The first image shows three adjacent segments that a merged together to provide spatial support for a sheep.
This image shows a grass region (not the boy) being composed of three regions.
I am making a project webpage which will shows results on the MSRC 23 Object Class dataset.
The first image shows three adjacent segments that a merged together to provide spatial support for a sheep.
This image shows a grass region (not the boy) being composed of three regions.
I am making a project webpage which will shows results on the MSRC 23 Object Class dataset.
Wednesday, June 06, 2007
a little bit more than computer vision?
Most object detection approaches in computer vision rely on computing features and a powerful classifier which predicts the presence of an object from some predefined object class such as bike, car, pedestrian, cat.
The 2007 PASCAL Visual Object Classes Challenge has 10 more object categories than the previous year. Three object categories that I would like to discuss are: 1.) chair, 2.) sofa, and 3.) tv/monitor.
Here are some images containing instances from those three object categories:
A problem arises when once considers the difference between a sofa and a chair. What is the visual difference between a large chair and a small sofa? Is it even worthwhile to engineer features to discriminate between these two object categories? Functionally and contextually, both are very similar object categories (things to sit on). The biggest difference between chairs and sofas is not only the size, but also the rooms they are located in. Unfortunately, is this something that an algorithm can learn by only training from instances of chairs and sofas?
Another interesting category in the PASCAL challenge is tv/monitor. For some reason a lapton screen is not considered an example of this category, and if an algorithm labeled the macbook monitor above as an instance of tv/monitor it would have been deemed incorred by the challenge! Perhaps the only way of determining that a screen is not a part of a laptop is by looking for the presence of a connected keyboard -- however, what is a disconnected keyboard is spatially close to a computer monitor? Are we expected to make such subtle distinctions?
What I'm trying to get at here, is that some of the object classes in the PASCAL 2007 challenge are ridiculous! While I do think that it is possible to detect these things in theory, the subtleties that are addressed in this challenge are simply beyond the realm of computer vision. If you consider the reasoning about object function, 3D layout, and context that is necessary to detect some of these "objects," then I do not expect any modern algorithm to do well for these object categories in 2007. Personally, I believe that the visual appearance of a "sofa" or "monitor" is not really very important in determining that it is that object. Once we realize that object recognition requires a little bit more than a powerful descriptor of the visual appearance, should we still treat object recognition as a part of computer vision? Computer metaphysics?
The 2007 PASCAL Visual Object Classes Challenge has 10 more object categories than the previous year. Three object categories that I would like to discuss are: 1.) chair, 2.) sofa, and 3.) tv/monitor.
Here are some images containing instances from those three object categories:
A problem arises when once considers the difference between a sofa and a chair. What is the visual difference between a large chair and a small sofa? Is it even worthwhile to engineer features to discriminate between these two object categories? Functionally and contextually, both are very similar object categories (things to sit on). The biggest difference between chairs and sofas is not only the size, but also the rooms they are located in. Unfortunately, is this something that an algorithm can learn by only training from instances of chairs and sofas?
Another interesting category in the PASCAL challenge is tv/monitor. For some reason a lapton screen is not considered an example of this category, and if an algorithm labeled the macbook monitor above as an instance of tv/monitor it would have been deemed incorred by the challenge! Perhaps the only way of determining that a screen is not a part of a laptop is by looking for the presence of a connected keyboard -- however, what is a disconnected keyboard is spatially close to a computer monitor? Are we expected to make such subtle distinctions?
What I'm trying to get at here, is that some of the object classes in the PASCAL 2007 challenge are ridiculous! While I do think that it is possible to detect these things in theory, the subtleties that are addressed in this challenge are simply beyond the realm of computer vision. If you consider the reasoning about object function, 3D layout, and context that is necessary to detect some of these "objects," then I do not expect any modern algorithm to do well for these object categories in 2007. Personally, I believe that the visual appearance of a "sofa" or "monitor" is not really very important in determining that it is that object. Once we realize that object recognition requires a little bit more than a powerful descriptor of the visual appearance, should we still treat object recognition as a part of computer vision? Computer metaphysics?
Monday, April 30, 2007
ipod won't turn on? -- try a gentle smack
Recently I've been having problems with my 4th Generation (ClickWheel) ipod turning on. It would just stop playing suddenly and then only display an apple support URL. My ipod wouldn't turn on (well only display the url). After a bit of googling, I found an answer!
I should first mention that everytime I would get the support url instead of the full ipod menu, I would be able to hear the ipod's harddisk whir. The noise was barely audible albeit unusually loud. Thus I figured that the problem was mechanical -- I've had the ipod for almost 3 years now.
To quote the solution I found in an online forum, "its fixed now - but I didnt use the updater. In sheer anger at it, and a hatred of ipods that developed because of this, I smacked it against the wooden desk I was at... then it made a funny whirring noise - hard disk? When I turned it on, it worked again. Who said violence isn;t the answer?"
I actually tapped the ipod against my palm and magic! The ipod worked. I didn't even have to perform surgery on my ipod like this guy.
I should first mention that everytime I would get the support url instead of the full ipod menu, I would be able to hear the ipod's harddisk whir. The noise was barely audible albeit unusually loud. Thus I figured that the problem was mechanical -- I've had the ipod for almost 3 years now.
To quote the solution I found in an online forum, "its fixed now - but I didnt use the updater. In sheer anger at it, and a hatred of ipods that developed because of this, I smacked it against the wooden desk I was at... then it made a funny whirring noise - hard disk? When I turned it on, it worked again. Who said violence isn;t the answer?"
I actually tapped the ipod against my palm and magic! The ipod worked. I didn't even have to perform surgery on my ipod like this guy.
Friday, March 09, 2007
Mundy-Style Vision and the Post-Midterm Generation
Vision has changed since 1987. Back in 1987, Joe Mundy gave a talk about Computer Vision and discussed the key of geometry role in image understanding understanding. Interestingly, he also gave some advice for students wanting to pursue computer vision. A nice quote from the video linked above can be found on the Wikipedia entry on Joseph Mundy, but the punchline is that he believed that (back in 1987) concepts such as optics, transformations, and algebraic surfaces need to be mastered in order to make advances in the field. Interestingly, twenty years later the vision of computer vision has taken a step in a different direction.
Pattern Recognition -- also known as machine learning, or simply put learning. I would say that if a student wants to pursue graduate study in Computer Vision they have a good background in Statistical Machine Learning. I will elaborate on this later.
On another note, I had my Statistical Machine Learning midterm this past week. Quite possibly my last in-class exam ever! I've been taking these things since middle school and it will be a nice change of pace. Only designing/grading midterms from now on.
Pattern Recognition -- also known as machine learning, or simply put learning. I would say that if a student wants to pursue graduate study in Computer Vision they have a good background in Statistical Machine Learning. I will elaborate on this later.
On another note, I had my Statistical Machine Learning midterm this past week. Quite possibly my last in-class exam ever! I've been taking these things since middle school and it will be a nice change of pace. Only designing/grading midterms from now on.
Monday, February 12, 2007
building vision systems like the good old days
Today, Alex Berg gave a VASC Seminar at CMU. At the end of the talk, Takeo Kanade made the interesting point that Alex's work on image parsing has a similar feel to the vision research in the late 70s. According to Takeo, in the good old days many PhD students would undertake the seemingly insurmountable task of building large vision systems and try to make it all work. More recently, PhD students are discouraged from such dreamer-style system-building and focus on more scientific study of narrower problems related to computer vision. In the past, students were actually discouraged from working on small 'easy' problems in vision and everybody tried to go for the gold.
An interesting question is: can we still build systems and be scientific about it (not simply program heuristics)? Alex speculated that vision will inevitably go back (is currently) to the system-building mentality due to the successess of using large amount of training data. I wonder how many things the new generation of PhD will re-invent.
An interesting question is: can we still build systems and be scientific about it (not simply program heuristics)? Alex speculated that vision will inevitably go back (is currently) to the system-building mentality due to the successess of using large amount of training data. I wonder how many things the new generation of PhD will re-invent.
Monday, January 22, 2007
Statistical Machine Learning
This semester I'm taking Statistical Machine Learning -- taught by John Lafferty and Larry Wasserman. Last year I sat in this class and decided that its worth taking for credit. The flipside is that I haven't taken one of the prerequisite classes, namely Intermediate Statistics, so I've been diligently studying so I can take the placement exam.
I've been studying from the Casella and Berger statistics books and its a great read!
I've been studying from the Casella and Berger statistics books and its a great read!
Wednesday, January 03, 2007
in park city!
I'm currently using iPass at a Starbucks in Park City, Utah. The mountains look amazing -- I'm buing a helment tonight and hitting the slopes early tomorrow.
As for today, I'm thinking nap, a steak dinner, and some much needed sleep.
As for today, I'm thinking nap, a steak dinner, and some much needed sleep.
going to park city, utah
I'm going to Park City, Utah to do some serious snowboarding in a few hours. I'm curious to see how my new sports gear (jacket/fleece/snowboard) perform. I'll post some pics in a few days.
Subscribe to:
Posts (Atom)