My checklist for making a "good" computer vision presentation
There are numerous articles about making or giving a good presentation. They mostly
talk about the general principles such as assessing the audience and be visual and active
during the presentation. Here are the specific points I believe helped me improve my
presentation skills.
- Average pictures (photos, figures, plots) per slide greater than 1.
If you can not include more than one picture to explain the concepts visually,
think twice, are you really doing computer vision? Strictly, no exceptions.
Visual demonstration is very effective, people can grasp a picture at a very high-level
in a flash, much faster than they can read text on the slide. As a result, they will
listen to what you say instead of what you is written on the slide. And, of course, they
can quickly re-refer to the picture in the verbal context you developed.
-
Include the jokes of different types.
Not everyone in the audience is a geek, but some of them are. Some of them are
interested in sports, a few of them are movie buffs, and many of them keep up with
the contemporary news. It is unlikely that everyone will enjoy a particular joke, but
including a couple of light moments satisfying one-subset-at-a-time may provide a
solution to the SET-COVER problem. Also, these light moments help the tired minds (read
people dozing off) catch with the rest.
- End the presentation with a funny pic.
Accept it, not everyone enjoyed the talk: some of them wanted more details, and some, fewer.
But everybody laughs at a good joke or a funny pic (George Bush has done a big service to the
community :D ). End the talk at a high note. Remember, the applaud, and hence the impression,
in the end is affected by
the final state of the mind the audience is left at. You may want them to ponder about the
intricate mathematics or the cool-ness of the approach after they have left the room, probably
not when you are still waiting for the happy response.
That said, I often find myself struggling with:
-
Reducing content
We like to talk, particularly, when it comes to talk about our own ideas, results, or
observations. We want everyone out there to know what is cooking in our brains. That is not
necessarily a very useful characteristic when it comes to giving a talk. As pointed out numerously
by my advisor and recently by Prof. Oliver Brock, a particular presentation could be more enjoyable
on average if some content or detail is simply dropped. Even though from your perspective, it
might be a very strong point, from the audience's perspective, it might be an extra baggage to carry.
And usually you want them to stroll, not scurry.
-
Time issues
Although I am not an advocate of practicing talks multiple times, keep in mind a rough estimate of
time per slide. In my experience, 1 min/slide is good enough, provided you have followed the
above-mentioned advice is also considered while preparing the presentation. The time management
depends on the obvious issues of the
speaker's normal speaking pace, the audience, and the specifics of the presentation.