2

Say I am interested in the latest development in the field of computer vision. I am a newbie into that area and do not have much knowledge about the history or the journey of computer vision till now or the latest development as well?

So what I do is usually go to google scholar and type some random combination of keywords and filter it based on years. I take the latest research paper that catches my eye (or take the latest review paper) and just start reading it. Usually when I am reading the introduction, I get references about the past development and start digging deeper.

This is where I get lost. I dig so deep into the references (rabbit hole of one reference to another) that by the time I am satisfied about what led me down that particular hole I have already lost track. This gets compounded as I do it for every interesting reference I find in the original paper.

How do I optimise my researching capacity?

My goal is to accomplish this from my researching capacity/ literature review ?

  1. Get important topics I can potentially look into (as I just select a topic that catches my eyes. I feel I am losing out on topics that may have more potential). How do you all select a topic you want to look into.

  2. Optimise the time required to form an opinion.

  3. Read literature quicker and more efficiently . (This down the rabbit hole technique I use gives me a lot of deep knowledge about a lot of aspects that I should not , my energy is expanded in an unoptimised way).

  4. Get to know prominent players in the field who provide high quality research that aligns with what I like to do.

  • 5
    Does this answer your question? How to read papers without falling into a rabbit hole? – GoodDeeds Aug 1 at 18:13
  • Not entirely. But it does give some good techniques. – Echo Aug 1 at 18:18
  • Do people write review article in this specific field? In my field (a branch of life sciences) many journals are dedicated exclusively to that, and reviews are the best way to have a broad perspective on a specific topic, based on expert-curated references. After reading three or four of those one can have a fairly good idea of what researchers have been up to in the past 5-10 years. – Mowgli Aug 3 at 2:17
3

Actually, I think you method isn't that bad in practice. I do something along the same lines. I'd just tinker a bit with your approach. For example:

  • Narrow your area. "Computer vision" sounds quite broad, focus on something more specific ("object recognition", "object tracking").
  • Pay attention to the citation count of a paper you are going to read and use it as a quick way to prioritize your reading.
  • If you opt to read highly-quoted papers only, it will greatly reduce your effort. You can read less-quoted papers later when you decide that you are interested in the specific problems they address. ("Less quoted" doesn't mean "bad", it might mean their authors study some narrow domain/problem, and the odds are that you don't need it at your current stage yet).
  • Use references in the papers you read as your "time machine" to uncover important papers you miss. Very often I see that the same reference constantly pops up here and there, but I miss it because it's too old to make it through my Google Scholar filter.
  • If you really want to get a broad view of a domain, read at least tables of contents of all major conferences/journals in the field for the past few years.

When I study something new, my first impression is always the same: "oh, lots and lots of papers, impossible to track and read them all; and all the problems are apparently solved here". However, in a week or so I start noticing that everything isn't so rosy, and there is a lot of work to be done, and many topics barely touched.

I must say though that "topics of higher potential" might be or might not be what you want to study. Consider, at least, your personal preferences (after all, you are going to spend much time working on a topic), and your existing skills/talents. It is possible that you can achieve more in a topic that might not be stellar, but fit nicely into your personality and background.

| improve this answer | |
  • Regarding your first point. What happens if I am not really sure about the sub domains of a broad domain. For instance, I just started looking into retinomorphic computer vision and I had a vague idea that I want to look into 'object recognition'. However, later I got to know there were other topics like 'pose etimation' that I completely missed cause I had no idea of its existence. How do I cope with such a situation. – Echo Aug 1 at 17:39
  • Yes, indeed, it might be not easy. I think a good "table of content" might help. For example, try Awesome lists. However, even there you'll need to dig quite deeply, but at least it's doable, because the space is finite. For example, "Pose estimation" is under "Computer Vision", "Awesome Deep Vision", "Human pose estimation". – rg_software Aug 1 at 17:47
  • Also do you know of any app where I can keep track of the papers I have read or how do you keep track of your progress ? Do you make a tree or how ? – Echo Aug 1 at 17:55
  • 2
    Oh, that's another question. It is highly dependent on your habits/preferences. Some people like mind maps, I don't. For now, I am okay with a reference manager (I use Citavi, but there are many similar systems). I place there full bib data and a PDF file of a paper. It allows to organize your thoughts in a tree form, but personally I don't do it. My very subjective belief is this: when you feel you need a way too complex organization, it means you have too much information, so something must be cut. – rg_software Aug 1 at 17:59
  • What's a quick way to find an exhaustive list of conferences and journals in my field? – FaureHu Aug 6 at 21:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.