A hybrid learning approach for better recognition of visual objects

Authors:
Ibrahim F. Imam;Srinivas Gutta
Affiliations:
SRA International, Fairfax, VA;Computer Science Department, George Mason University, Fairfax, VA
Venue:
AAAI'96 Proceedings of the thirteenth national conference on Artificial intelligence - Volume 2
Year:
1996

Citing 5
Cited 0

Synergy of clustering multiple back propagation networks

Advances in neural information processing systems 2
The Meta-Pi Network: Building Distributed Knowledge Representations for Robust Multisource Pattern Recognition

IEEE Transactions on Pattern Analysis and Machine Intelligence
Learning decision trees from decision rules: a method and initial results from a comparative study

Journal of Intelligent Information Systems - Special issue on methodologies for intelligent systems
A Survey of Glove-based Input

IEEE Computer Graphics and Applications
Democracy in neural nets: voting schemes for classification

Neural Networks

Quantified Score

Hi-index	0.01

Visualization

Abstract

Real world images often contain similar objects but with different rotations, noise, or other visual alterations. Vision systems should be able to recognize objects regardless of these visual alterations. This paper presents a novel approach for learning optimized structures of classifiers for recognizing visual objects regardless of certain types of visual alterations. The approach consists of two phases. The first phase is concerned with learning classifications of a set of standard and altered objects. The second phase is concerned with discovering an optimized structure of classifiers for recognizing objects from unseen images. This paper presents an application of this approach to a domain of 15 classes of hand gestures. The experimental results show significant improvement in the recognition rate rather than using a single classifier or multiple classifiers with thresholds.