[cs-talks] Updated info - Upcoming PhD Thesis (Mon- Jul 18)
cs at bu.edu
Tue Jul 12 09:26:42 EDT 2016
PhD Thesis Proposal Defense
3D Pose Estimation of Flying Animals in Multi-View Video Datasets
Mikhail Breslav, BU
Monday, July 18, 2016 at 1:00pm in MCS 148
Abstract: Flying animals such as bats, birds, and moths are actively studied by researchers wanting to better understand these animals' behavior and flight characteristics. Towards this goal, multi-view videos of flying animals have been recorded both in laboratory conditions and natural habitats. The analysis of these videos has shifted over time from manual inspection by scientists to more automated and quantitative approaches based on computer vision algorithms.
In this thesis we study the largely unexplored problem of 3D pose estimation of flying animals in multi-view video data. This problem has received little attention in the computer vision community where few flying animal datasets exist. Additionally, published solutions from researchers in the natural sciences have not taken full advantage of advancements in computer vision research. Our thesis addresses this gap by proposing three different approaches for 3D pose estimation of flying animals in multi-view video datasets, which evolve from successful pose estimation paradigms used in computer vision. Our first approach models the appearance of a flying animal with a synthetic 3D graphics model and then uses a Markov Random Field to model 3D pose estimation over time as a single optimization problem. Our second approach builds on the success of Pictorial Structures models and further improves them by automatically discovering parts from regions of training images without annotations. Our method uses the discovered parts to generate more accurate appearance likelihood terms which in turn produce more accurate landmark localizations. Our third approach takes advantage of the success of deep learning models and adapts existing deep architectures to perform landmark localization. Both our second and third approaches perform 3D pose estimation by first obtaining accurate localization of key landmarks in individual views, and then using calibrated cameras and camera geometry to reconstruct the 3D position of key landmarks.
We show that our proposed algorithms generate first-of-a-kind and leading results on real world datasets of bats and moths, respectively. Furthermore, we make a variety of resources freely available to the public to further strengthen the connection between research communities.
Margaret Betke, Stanley Sclaroff, Tyson Hedrick, Wayne Snyder, and Kate Saenko
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cs-talks