Scientific Imaging Research Group

Blog post edited by Anonymous - "Migrated to Confluence 4.0"

Historically, experimental observations in science have been limited to handwritten logs and field diaries. However, the recent influx of low cost digital cameras allows researchers who rely on visual observations to digitally record experiments. This has increased the amount of data available to researchers and allows researchers to time-shift their observational processes, so that multiple cameras can continually record experiments and researchers can review and re-review the data at their leisure. Such low-cost cameras can gather a tremendous amount of digital data, but there is no simple, automated method for examining this data and extracting the information necessary for scientific measurements. Thus, researchers typically use man-hours (hire students) to manually annotate video data frame-by-frame, which is an extremely slow process subject to variations in quality and detail.

The problem is that the amount of data produced by existing digital cameras is many orders of magnitude larger than the scientific observations needed by the researchers. Our research team is developing new methodologies to facilitate the scientific process and make it more affordable to filter large amounts of image data into observations that can be used to test research hypotheses. The goal of this research is to find ways to improve scientific workflow so that researchers can scale-up faster and minimize their mean time to science.

This summer, we are going to focus on two domains centered around image phenomics:

• Working with Dr. Fred Dyer from Zoology to extend prior work we have done to develop ChamView.
• Working with Dr. Ian Dworkin from Zoology to try to automate his work with images of fly wings.

As part of these projects, we are also collaborating with Dr. Yang Wang from Mathematics and Dr. Lifeng Weng from Statistics, who are helping to develop new learning algorithms for these image phenomics projects.

• Dirk

View Online

Blogpost migrated from ICER Wiki using custom python script. Comment on errors below.