• Best reading practices and resources for research can be found out organically as you start, through (sometimes painful) trial and error cycles, but there must be a better way !

• I am a PhD student in Computer Vision, in the deep-learning era, my reading can be roughly separated in two categories:

  • deep learning papers with usually a marginal theoretical input but important experimental findings which provide building-blocks that might be useful to me
  • domain-specific papers that depict methods to produce state-of-the-art results on some task

• While reading I often need both understand the general idea behind a paper as well as low-level implementation details, but I rarely need to understand complex math equations

Computer vision has attracted significant interest since deep-learning has made its latest breakthrough in 2012, this means the field gets more and more crowded, and keeping up with latest papers can become a challenge.

Thankfully there are some tools that can help you with that !

• Other people have gone through the trial-and-error process before you, and maybe instead of repeating the process you can learn from them
• After going through several "How did I not find out about this earlier?!" moments, I think it would be great to gather in writing some tools and best-practices that I wish I had found out about earlier !

Is your research similar to mine ? Could this be useful to you ?

• I am a PhD student in Computer Vision, and most of my research involves coding and running experiments (little equations, a lot of visualizations, much more time spent in the terminal then with a pen and a paper)
• My remarks are therefore mostly focused on optimizing experiment-centric research (in contrast to, for instance, theoretical research, for which I have no previous experience !)