Would a question asking for a list of good computer vision text books be on-topic? There is not a single "correct" answer here, but the answers can be ranked by votes.

EDIT: Let me expand a bit. I have taught a graduate computer vision course last fall, and I had a hard time selecting a textbook. Getting a list of suggestions ranked by votes from people who have taken or taught a vision course would have been very helpful.


3 Answers 3


I say it is, given that the person gives a precise description of what it is they'd like to learn. I may be wrong (and someone please correct me if I am), but the tag is precisely for these purposes.

If you ask "What's a good book on DSP?" you'll get downvoted by 30 people. If you ask "Does anyone know a good paper or book that describes applications of Empirical Mode Decomposition to EEG signals?", I'm sure this is valid.


While I can see the value in book recommendation questions, such questions are a terrible fit for the StackExchange platform, mainly because "letting good answers bubble up to the top" doesn't work very well with lists.

For example, if you were to ask "What is a good beginner's book for learning DSP?", invariably, one of the answers, if not the first, will be "Discrete time signal processing" by Oppenheim and Schafer. Now I cannot possibly fathom of a DSP site where this answer would not receive copious upvotes. Does this really answer the question? While it may be a great book, is it truly a good book for a beginner? Who is a beginner? This is a soft description that is inherently dependent on the individual's assessment of themselves and what is a beginner book for one might not be for the other. Now let's say that a month later, you added an answer linking to an amazing book that could teach your grandma DSP in two days. Would it really be able to catch up with the first answer? This is where such list of X and recommendation questions break down.

However, I agree that this is a very specialized site with a narrow focus and it is indeed hard to find good recommendations from a simple Google search (unlike searching for books that teach C++). Hence such questions could greatly add to the value of the site, if we have a system in place that wouldn't let the answers be crapped upon by link droppers.

I suggest that requests/book requests be allowed, with the following conditions:

  • Questions should be specific and to the point. For example, "What book is good for DSP" should be nuked outright, whereas "What is a good reference for the theory of adaptive filters?" or "What is a good intro book for learning particle filters" is allowed.
  • Answers should be accompanied by why the OP thinks it is a good book (personal experience required). Mere links or regurgitating Amazon reviews is insufficient and such answers should be either nuked/moved to a comment.
  • If you know of a canonical/seminal book, but don't have experience with it, it should be a comment.

This can be worked upon/tweaked, but some strict guidelines are necessary.

  • 1
    $\begingroup$ Based on my experience at electronics.stackexchange, you'll have to be extremely selective. There are a lot of programmers on Stackoverflow who want to branch out their experience in programming-related fields, and they will come by and ask "How do I get started" questions. If we're going to allow these, we need to be ruthless in enforcing the conditions. $\endgroup$ Commented Sep 22, 2011 at 11:38

Please take a look at a similar question asked at Programmers.SE.

The usefulness of the StackExchange family of sites is that most question seekers will be able to get an answer (or some clues) from reading a post and its responses, without having to look further. Thus there are two requirements:

  • The question should ask for a source or reference for a narrow and specific theory or technique.
    • An example of a topic that is too broad is one that would require a whole graduate-level course to cover.
  • The answer should include a brief excerpt that gives the reader some indication about what type of information will be provided by the book (the level of detail, mathematical treatment, history, sample code etc.)

A book recommendation for one general topic (such as "computer vision") could be added to the tag wiki. Here is a similar proposal on Programmers.SE. (Disclaimer: that question was asked by me.)

Judged by the comments left at the affiliated book-seller website, I'm afraid that letting users vote on book recommendations might produce polarized results.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .