[Work Log] Work Log

January 07, 2014

Layout organization, pick a section, start writing.

Proposal guide, CS department, Ohio University [pdf]

In researching multiple-view plant datasets, I came across the LifeCLEF 2014 challenge. The goal is plant-species identification from single images. Images are of several different organs (leafs, fruit, stems, branches) or entire structures (branches, entire plant). Most images are in the wild, but some (leafs) are from a flatbed scanner. All images are given a one-to-five rating based on their quality. Even this isn't the type of task we're focusing on, it's valuable to note what types of tasks the community is focusing on.

Also available is a fish classification task from videos. It might be interesting to try to track GP curves in an image and classify fish based on their GP parameters (temporal and spatial). The cubic spline model is already scale-invariant, so scale issues might not be a big deal. GP-LVM model might be even more interesting.

TODO - look into CVPR 2013 "thin branch" paper. Existing datasets? Their dataset? Literature reivew?

Posted by Kyle Simek
blog comments powered by Disqus