2013/02/19 by Walter J. Scheirer, Michael J. Wilber, Scheirer, Walter J. +5 · 1 citation
Computer Science · Mathematics · #Advanced Image and Video Retrieval Techniques #Artificial intelligence #Computer Vision and Pattern Recognition (cs.CV) #Computer science #Context (archaeology) #Domain Adaptation and Few-Shot Learning #FOS: Computer and information sciences #Face and Expression Recognition #Leverage (statistics) #Machine learning #Matching (statistics) #Mathematics #Metric (unit) #Pattern recognition (psychology) #Theoretical computer science #cs.CV
paper · pdf · doi:10.48550/arxiv.1302.4673
published in arXiv (Cornell University) (Cornell University) · 9 pages, 5 figures
arxiv created 2013/02/19 · openalex publication_date 2013/02/19 · arxiv updated 2013/02/20 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/05
Recognition is the fundamental task of visual cognition, yet how to formalize the general recognition problem for computer vision remains an open issue. The problem is sometimes reduced to the simplest case of recognizing matching pairs, often structured to allow for metric constraints. However, visual recognition is broader than just pair matching -- especially when we consider multi-class training data and large sets of features in a learning context. What we learn and how we learn it has important implications for effective algorithms. In this paper, we reconsider the assumption of recognition as a pair matching test, and introduce a new formal definition that captures the broader context of the problem. Through a meta-analysis and an experimental assessment of the top algorithms on popular data sets, we gain a sense of how often metric properties are violated by good recognition algorithms. By studying these violations, useful insights come to light: we make the case that locally metric algorithms should leverage outside information to solve the general recognition problem.