The self-organizing hierarchical optimal subspace learning and inference framework for view-based object recognition and image retrieval
D.L. Swets · Michigan State University Libraries · 1996
A self-organizing framework for object recognition is described. The SHOSLIF (Self-Organizing Hierarchical Optimal Subspace Learning and Inference Framework) system uses the theories of optimal linear projection for automatic optimal feature selection and a hierarchical structure to achieve a logarithmic retrieval complexity. A Space-Tessellation Tree is automatically generated using the Most Expressive Features (MEFs) and the Most Discriminating Features (MDFs) at each level of the tree. We use incremental learning for large-scale and open-ended learning problems, and we allow for perturbations in the size and position of objects in the images through learning. We demonstrate the technique of a large database of widely varying real-world objects in natural settings, and show the applicability of the approach even for large variability within a particular class, including 3D rotation.