On May 18th, I gave a presentation at our local Kaggle Meetup group on a recent Kaggle competition.

The competition was put forth by The Nature Conservancy. The task was to automatically detect and identify fish from still images taken from cameras mounted on fishing vessels. This would help in the enormous effort of monitoring fishing activity in sensitive areas.

I competed on a team (Fishy RSeaNN) for this competition. We placed 84th out of 2293 teams. This put us (barely) in the top 4% of the competition, but we still performed worse than the sample submission for the competition.

This presentation discusses the competition and its challenges, some winning strategies used by the top-10 winners of the competition and some state of the art techniques in object detection and classification within images. If you want to follow along with the slides, you can find them here.

(Apologies for the number of “umms” in this video. This is a great example of why you should record yourself giving presentations so that you can hear yourself and your presentation style before you present in front of other people. Regardless, if you can get past the filler, the content is pretty good. Enjoy!)