Mean Shift Is a Bound Optimization
IEEE Transactions on Pattern Analysis and Machine Intelligence
Recovering 3D Human Body Configurations Using Shape Contexts
IEEE Transactions on Pattern Analysis and Machine Intelligence
Building Models of Animals from Video
IEEE Transactions on Pattern Analysis and Machine Intelligence
Proceedings of the 6th ACM international conference on Image and video retrieval
Learning Layered Motion Segmentations of Video
International Journal of Computer Vision
IEEE Transactions on Image Processing
Region covariance matrix-based object tracking with occlusions handling
ICCVG'10 Proceedings of the 2010 international conference on Computer vision and graphics: Part I
CVPR'04 Proceedings of the 2004 IEEE computer society conference on Computer vision and pattern recognition
PR'05 Proceedings of the 27th DAGM conference on Pattern Recognition
Learning hierarchical shape models from examples
EMMCVPR'05 Proceedings of the 5th international conference on Energy Minimization Methods in Computer Vision and Pattern Recognition
Motion Coherent Tracking Using Multi-label MRF Optimization
International Journal of Computer Vision
Background modeling methods for visual detection of maritime targets
Proceedings of the 4th ACM/IEEE international workshop on Analysis and retrieval of tracked events and motion in imagery stream
Hi-index | 0.00 |
This paper describes a system that can build appearance models ofanimals automatically from a video sequence of the relevant animalwith no explicit supervisory information. The video sequence neednot have any form of special background. Animals are modeled as a2D kinematic chain of rectangular segments, where the number ofsegments and the topology of the chain are unknown. The systemdetects possible segments, clusters segments whose appearance iscoherent over time, and then builds a spatial model of such segmentclusters. The resulting representation of the spatial configurationof the animal in each frame can be seen either as a track - inwhich case the system described should be viewed as a generalizedtracker, that is capable of modeling objects while tracking them -or as the source of an appearance model which can be used to builddetectors for the particular animal. This is because knowing avideo sequence is temporally coherent - i.e. that a particularanimal is present through the sequence - is a strong supervisorysignal. The method is shown to be successful as a tracker on videosequences of real scenes showing three different animals. For thesame reason it is successful as a tracker, the method results indetectors that can be used to find each animal fairly reliablywithin the Corel collection of images.