sher@rochester.UUCP (David Sher) (09/02/85)
A short time ago there was a discussion of whether there should be a group for discussion of image processing/computer vision such as net.ai.vision. I don't know if this discussion occured just in net.graphics or also in net.news.group (I don't read net.news.group). Anyway there were lots of people interested in a group. Right now such discussion is going on in net.graphics. This results in me n'ing about 75-90% of the articles in order to separate out the articles that pertain to my research! While many graphics people are interested in computer vision many computer vision researchers and experts are not interested in graphics! (me for one). Anyway since it is a field that has hundreds of researchers on the net and many people are interested in it I suggest that the issue be reopened about whether net.ai.vision should exist. Unfortunately with the term starting I don't know if I can count on being able to survey and count the votes. I also lack the divine attributes necessary to generate such a newsgroup. -- -David Sher sher@rochester seismo!rochester!sher
ken@turtlevax.UUCP (Ken Turkowski) (09/05/85)
In article <11307@rochester.UUCP> sher@rochester.UUCP (David Sher) writes: >A short time ago there was a discussion of whether there should >be a group for discussion of image processing/computer vision such as >net.ai.vision. ... Right now such discussion >is going on in net.graphics. This results in me n'ing about 75-90% >of the articles in order to separate out the articles that pertain to >my research! When you read through the printed literature on pattern recognition, don't you also have to weed through articles not pertaining to your research? >While many graphics people are interested in computer vision >many computer vision researchers and experts are not interested in graphics! >(me for one). Anyway since it is a field that has hundreds of researchers >on the net and many people are interested in it I suggest that the >issue be reopened about whether net.ai.vision should exist. It is unclear that a subgroup of AI would be the proper group. Part of vision is concerned with image processing, the other with pattern recognition. The mathematics associated with these is substantially dissimilar that such discussions may not belong in the same group. The image processing portion is a subset of signal processing, and who knows what kind of mathematics is used for pattern recognition (fuzzy set theory? statistics?). How about net.sigproc or net.signal for discussions about signal processing? This would include things like analog and digital filters, stability, anti-aliasing, image processing, Kalman filters, parameter estimation, correlation, signal theory, source coding, etc. What sort of interest is there in such topics? What techniques are used in computer vision? What are the current hot topics? Who knows? Who cares? Of those who care, what are your other interests? What sort of catagories make sense to interest a critical mass of people? Or should there be a mailing list set up for such things? Kahn at UCLA has such a mailing list for ARPA-type people, but it has been VERY inactive. Lets have some discussion! David Sher, are you there? How about other researchers and hackers? -- Ken Turkowski @ CADLINC, Menlo Park, CA UUCP: {amd,decwrl,hplabs,seismo,spar}!turtlevax!ken ARPA: turtlevax!ken@DECWRL.ARPA
sher@rochester.UUCP (David Sher) (09/07/85)
> > When you read through the printed literature on pattern recognition, > don't you also have to weed through articles not pertaining to your > research? I never have to avoid billions of articles on graphics, I don't get the graphics literature because of this. Most of the pattern recognition literature has some application to my studies however. > Ken here discusses pattern recognition vs image processing > There is a fine line between patter recognition and image processing. Much of image processing implicitly is involved in recognizing patterns in images and enhancing them. Image enhancement can ease the job of a pattern recognizer. As far as the mathematics of pattern recognition it is a hodgepodge of signal detection theory, information theory, biological models, and statistical inference. (I probably missed some). > > How about net.sigproc or net.signal for discussions about signal processing? > This would include things like > analog and digital filters, X > stability, ? > anti-aliasing, * > image processing, * > Kalman filters, ? Whats a Kalman filter? What is it for? (I am a converted AI/Logic person and don't have a very complete knowledge of signal processing. (I will by the time I finish my thesis though!)) (feel free to mail me if not of general interest) > parameter estimation, * > correlation, * > signal theory, * > source coding, etc. X > > What sort of interest is there in such topics? What techniques are > used in computer vision? What are the current hot topics? Who knows? > Who cares? Of those who care, what are your other interests? What > sort of catagories make sense to interest a critical mass of people? > I have anotated the list of topics above according to what is relevant to my picture of computer vision. * means that it is relevant, X means that it is irrelevant, and ? means I don't know what is refered to. Some hot topics in computer vision are: reconstructing the 3D scene from a 2D image : Subtopics shape from shading shape from optical flow deriving optical flow from images 3D motion from optical flow 3D motion from a sparse set of correlated points Evidence theory (combining the output of unreliable operators to get reliable results) Segmentation Object recognition Passive Navigation Models of the human visual system These are the kinds of topics I have come in contact with here at the University of Rochester. I will be glad to post short explanatory notes (at my convinience of course) explaining the meaning of some of these obscure names. There is no doubt much work in computer vision that is not done according to these topics so if anyone has any additions let me know. > Or should there be a mailing list set up for such things? Kahn at UCLA > has such a mailing list for ARPA-type people, but it has been VERY > inactive. Lets have some discussion! David Sher, are you there? How > about other researchers and hackers? Possibly the reason for the inactivity has been poor publicity. I just heard about it in response to my message! I think one can send to this group by mailing to vision@aids-unix but am not sure this is correct. I will probably soon send a message to net.graphics and vision@aids-unix about scene models. Oh yes, I am here but should be working on my thesis! > -- > Ken Turkowski @ CADLINC, Menlo Park, CA > UUCP: {amd,decwrl,hplabs,seismo,spar}!turtlevax!ken > ARPA: turtlevax!ken@DECWRL.ARPA *** REPLACE THIS LINE WITH YOUR FOOLISHNESS *** -- -David Sher sher@rochester seismo!rochester!sher
dak2@mtuxt.UUCP (D.KALL) (09/11/85)
Yes David I agree. One vote for. Who do we contact to get one?