On Friday my students Joseph Vargovich and Atiyeh Fotoohinasab will present at a local poster session for students of the School of Informatics, Computing, and Cyber Systems. Crystal Hepp recommended that they use the poster design described in How to create a better research poster in less time, which is a 20 minute video describing the problems with current mainstream poster designs/sessions, and some solutions. Overall I think the video has some interesting ideas but there are a few points that I would like to comment on.

• in the video the poster is thrown away after the session. Instead, I usually bring the poster home and put it up in my lab for discussion with colleagues/labmates.
• the video mentions that legible text is important – I agree! Especially make sure that all figure/axis/legend text is readable.
• in the video the grad student gets a hand me down powerpoint template! PowerPoint is for presentations, not poster design. Instead, use Scribus or Inkscape. Scribus is preferable if you are still working on the figures, because it supports including linked figure files (which are automatically updated in the poster if you update the source figure file). Here is the scribus sla file that I used for my NIPS’14 PeakSeg poster.
• make it easy/fast? I guess this is a good goal, but there is no substitute for hard work / dedicated time working on your poster.
• the video suggests color coded poster types, e.g. green for intervention, red for theory, etc. Conference organizers may suggest the color scheme they would like to encourage.
• displaying the main result as big text in middle seems to be effective. I would suggest adding one or two main result figures in the middle below that.
• the “ammo bar” on the right should have most of the other figures from the paper.
• the “silent presenter” bar on the left contains the main title, which is a more technical version of the main result.
• qr code should be 5 inches! That seems pretty large to me.
• No commentary in the video about how to display actual computer code / pseudocode for algorithms / equations, which are actually important in machine learning and statistical software.
• No mention about figure file formats. For figures with equations I use tikzDevice (tikz function in R) to generate a tex file which the pdflatex command line program then converts to PDF for final inclusion in the poster. For figures without equations you should use either vector graphics (pdf/svg functions in R) or high-resolution raster graphics, e.g. png("figure-1.png", width=5, height=5, units="in", res=1000) in R. GIMP can be used to create high-resolution raster graphics from vector graphics such as PDF (the desired resolution can be specified along with the page number when the PDF file is first opened/imported).