9.67 | Spring 2001 | Undergraduate

Object and Face Recognition

Study Materials

Experimental Studies of Recognition

  • Pattern recognition by invertebrates
    1. How do honey bees recognize shapes?
    2. How do bees and ants use their recognition abilities to navigate?
  • Pattern recognition by invertebrates
    1. How do octopii recognize shapes?
  • Pattern recognition by robots modeled on invertebrates

Student Presentation

How can we enable a robot to navigate based on insect recognition strategies?

  • Pattern recognition by birds
    1. Can birds recognize and categorize complex shapes/scenes?
    2. What recognition strategies do birds use?
  • Pattern recognition by birds
    1. Recognition of individuals in a gull colony
    2. Recognition by chicks of mother’s beak via key-signs
    3. Pigeons’ recognition of Monet and Picasso paintings
  • Object recognition by humans (psychophysics)
    1. What are the key characteristics of human recognition performance (speed, view-point/orientation dependency, reliance on prototypes…)?
  • Object recognition by humans (psychophysics)

Student Presentation

Can human observers recognize highly impoverished motion sequences?

  • Object recognition by primates (physiology)
    1. Does the brain have areas specialized for recognition?
    2. Are there different functional streams in the brain for recognition and spatial analysis?
  • Object recognition by primates (physiology)
    1. What have functional imaging studies told us about brain Mechanisms of recognition?
  • Object recognition deficits following brain damage in primates
    1. What have lesion studies in monkeys told us about brain mechanisms of recognition?
    2. What are the basic characteristics of visual agnosias and how are they correlated with the nature of damage?
  • Object recognition deficits following brain damage in primates

Student Presentation

A case study of visual agnosia (‘The man who mistook his wife for a hat’)

  • Beyond individual object recognition
    1. How do humans recognize large scenes?
    2. Does scene-context influence individual object recognition? Can we formalize a model of contextual influences?
    3. What is the role of eye- movements in scene perception?
  • Beyond individual object recognition

Student Presentation

How are scenes encoded in memory? - studies using the ‘change-blindness’ paradigm

  • Development of object perception
    1. Can babies parse the visual world into objects?
    2. How and when do babies acquire knowledge of object properties?
  • Development of object perception

Student Presentation

Case study of sight recovery in adulthood (“To see and not see,” Oliver Sacks)

Computational Studies of Recognition

  • Classical pattern classification theory
    1. Bayes decision theory
    2. Supervised and unsupervised learning
  • Classical pattern classification theory

Student Presentation

Case study of statistical pattern classification: A trainable tool for finding small volcanoes in SAR Imagery of Venus

  • Computational theories of object recognition
    1. Theories based on 3D object models
    2. Theories based on 2D image models (alignment approach; linear combination of views)
  • Computational theories of object recognition

Student Presentation

  • Case study- Brook’s ACRONYM system of recognition based on 3D models and symbolic reasoning
  • Computational theories of object recognition

Student Presentation

Using linear object models for recognition

  • Image and model correspondence
    1. How can we determine the matching features in images and models?
    2. How can we segment images into objects and objects into parts?
  • Image and model correspondence

Student Presentation

Segmentation via saliency computations

  • Network models of object recognition
    1. Feedforward models of recognition (Fukushima, RBFs)
    2. Feedback models of recognition (Ullman)
  • Network models of object recognition

Student Presentation

A particular network model of recognition - Mumford’s scheme

  • Notable case studies of artificial recognition schemes
    1. The first artificial recognition system (Roberts)
    2. Histogram based recognition (Swain and Ballard)

Face Recognition

  • Face recognition vs. general object recognition
    Are faces special? (Evidence from physiology, neuropsychology, Psychophysics, imaging and developmental studies)
  • Face recognition vs. general object recognition

Student Presentation

Are faces special? Psychophysics and imaging with dog experts

  • Face recognition studies
    1. Is face recognition feature-based or holistic?
    2. What are the salient shape and surface cues in a face?
  • Face recognition studies

Student Presentation

What can facial caricatures tell us about face recognition processes?

  • Social aspects of face recognition
    How do we perceive facial affect, gaze direction, and aesthetics?
  • Social aspects of face recognition

Student Presentation

Do babies prefer attractive faces?

  • Psycho- forensic aspects of face recognition
    1. What are the forensic applications of facial recognition research?
    2. Can people be trained to be better encoders of faces?
  • Psycho- forensic aspects of face recognition

Student Presentation

  1. What are the current facial composite creation systems?
  2. Can they be improved based on research results?
  • Psycho- forensic aspects of face recognition
  • A demonstration of the IdentiKit system by a local police artist
  • Implemented systems for face recognition
    An examination of a few notable systems:
    1. Turk’s Eigenface based system
    2. Von der Marlsburg’s graph based system
    3. Beymer’s template based system
  • Implemented systems for face recognition
    Other biometric systems:
    1. Iris recognition
    2. Retina recognition
    3. Face recognition in IR
    4. Ear recognition

Synthesis and Open Issues

  • Closing the loop between recognition and perception
    1. Can recognition influence early perception? - historical ideas
    2. Is there any experimental evidence to support this idea?
  • Grand synthesis
    1. Might different sensory modalities share similar recognition strategies?
    2. What are the key open questions in the area of recognition?
  • Object recognition research at MIT
    What are the opportunities for research in high-level vision in BCS and the AI lab?

Course Info

Instructor
As Taught In
Spring 2001
Learning Resource Types
Written Assignments
Activity Assignments