Scaling up biologically-inspired computer vision: A case study in unconstrained face recognition on facebook
Authors:
- 1 June 2011
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 21607508,p. 35-42
Abstract
Biological visual systems are currently unrivaled by artificial systems in their ability to recognize faces and objects in highly variable and cluttered real-world environments. Biologically-inspired computer vision systems seek to capture key aspects of the computational architecture of the brain, and such approaches have proven successful across a range of standard object and face recognition tasks (e.g. [23, 8, 9, 18]). Here, we explore the effectiveness of these algorithms on a large-scale unconstrained real-world face recognition problem based on images taken from the Face-book social networking website. In particular, we use a family of biologically-inspired models derived from a high-throughput feature search paradigm [19, 15] to tackle a face identification task with up to one hundred individuals (a number that approaches the reasonable size of real-world social networks). We show that these models yield high levels of face-identification performance even when large numbers of individuals are considered; this performance increases steadily as more examples are used, and the models outperform a state-of-the-art commercial face recognition system. Finally, we discuss current limitations and future opportunities associated with datasets such as these, and we argue that careful creation of large sets is an important future direction.Keywords
This publication has 16 references indexed in Scilit:
- A High-Throughput Screening Approach to Discovering Good Forms of Biologically Inspired Visual RepresentationPLOS Computational Biology, 2009
- What is the best multi-stage architecture for object recognition?2009
- Attribute and simile classifiers for face verification2009
- Multiple One-Shots for Utilizing Class Label Information2009
- What drives content tagging2008
- Why is Real-World Visual Object Recognition Hard?PLOS Computational Biology, 2008
- Untangling invariant object recognitionTrends in Cognitive Sciences, 2007
- HT06, tagging paper, taxonomy, Flickr, academic article, to read2006
- Hello! My name is... Buffy'' -- Automatic Naming of Characters in TV Video2006
- Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in positionBiological Cybernetics, 1980