|Course Dates||Length||Meeting Times||Status||Format||Instructor(s)||CRN|
|June 28, 2021 - August 11, 20216/28 - 8/11||6 Weeks||Online||Open||Online||Chen Ming||11743|
Artificial Intelligence has the possibility to impact our daily lives in many ways. One lesser known application is to help behavioral scientists reevaluate how they think about and design experiments. Similar to how you know what your cat, dog or other pet needs by observing their behavior a computer can be trained to observe, identify and catalog different animal behaviors in a laboratory setting. Even recently, scientists spent months analyzing hundreds of hours of videos recorded in experiments, frame by frame, with the help of an army of undergrads to answer scientific questions about behavior. But now a trained artificial neural network using deep learning techniques can do video analysis in no time. Researchers can now focus on answering the underlying scientific questions, instead of tedious data extraction. They may even be overwhelmed by the amount of data artificial intelligence will provide them.
This course will not only offer an overview of various applications of deep learning in neuroethological research, but also allow students to track subjects in videos on their own with deep learning tools in hands-on sessions. Though it may sound intimidating, graphical user interfaces designed by the creators have largely reduced the difficulty for the less computer-savvy to take advantage of deep learning programs. The instructor of this course will provide the students with step by step instructions for analyzing and conducting this work.
Tracking the body parts of subjects in an experiment is the basis to understand animal behaviors and their neural response. Just a few years ago, posture estimation could take a long time, starting from training the subject to be familiar with markers, manual labeling, and then tracking the body parts. But now with Deeplabcut, a deep learning program that can estimate the posture without markers on the subject, highly efficient tracking can be realized.
Students will be engaged with minimal fundamentals of artificial intelligence and deep learning. Then students will learn how to track a single subject in videos using Deeplabcut, and groups of animals using idtracker.ai. Programming languages involved in the course will be introduced in the beginning. Students will practice their skills learned in this class by completing a final project on their own, where they will find or shoot their own videos and try to track certain object in the videos.
Upon completing this course, students will get deeper understanding of the interdisciplinary trends in higher education, where any discipline is welcome to seek inspiration and help from other disciplines. They will also have learned what artificial intelligence is, what it can do, and how to exploit it as a tool for their future research projects in behavioral studies or other disciplines. The two disciplines included in this course – computer science and neuroethology – will help students with their choice of major at colleges as well.
Prerequisites: This course is open to rising juniors and seniors who are familiar with basic computer operations.