Egocentric visual event classification with location-based priors

Authors:
Sudeep Sundaram;Walterio W. Mayol-Cuevas
Affiliations:
Department of Computer Science, University of Bristol;Department of Computer Science, University of Bristol
Venue:
ISVC'10 Proceedings of the 6th international conference on Advances in visual computing - Volume Part II
Year:
2010

Citing 8
Cited 1

Realtime Personal Positioning System for Wearable Computers

ISWC '99 Proceedings of the 3rd IEEE International Symposium on Wearable Computers
Recognizing User Context via Wearable Sensors

ISWC '00 Proceedings of the 4th IEEE International Symposium on Wearable Computers
Space-time Interest Points

ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
Histograms of Oriented Gradients for Human Detection

CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 1 - Volume 01
Learning and inferring transportation routines

Artificial Intelligence
A Linear Programming Approach to Max-Sum Problem: A Review

IEEE Transactions on Pattern Analysis and Machine Intelligence
Video-rate localization in multiple maps for wearable augmented reality

ISWC '08 Proceedings of the 2008 12th IEEE International Symposium on Wearable Computers
When Else Did This Happen? Efficient Subsequence Representation and Matching for Wearable Activity Data

ISWC '09 Proceedings of the 2009 International Symposium on Wearable Computers

Hierarchical Hidden Markov Model in detecting activities of daily living in wearable videos for studies of dementia

Multimedia Tools and Applications

Quantified Score

Hi-index	0.00

Visualization

Abstract

We present a method for visual classification of actions and events captured from an egocentric point of view. The method tackles the challenge of a moving camera by creating deformable graph models for classification of actions. Action models are learned from low resolution, roughly stabilized difference images acquired using a single monocular camera. In parallel, raw images from the camera are used to estimate the user's location using a visual Simultaneous Localization and Mapping (SLAM) system. Action-location priors, learned using a labeled set of locations, further aid action classification and bring events into context. We present results on a dataset collected within a cluttered environment, consisting of routine manipulations performed on objects without tags.