Visual object recognition: do we know more now than we did 20 years ago?

Area of Science:

• Cognitive Science
• Neuroscience
• Computer Vision

Background:

• Object recognition research has progressed over two decades, evolving from early structural-description models.
• Psychophysical evidence and neurophysiological studies of the visual cortex have challenged and refined recognition theories.
• Advancements in neuroimaging, particularly functional magnetic resonance imaging (fMRI), have provided new insights.

Purpose of the Study:

• To review the historical progress and key developments in the field of object recognition.
• To highlight the shift from structural-description to view-based models.
• To discuss the impact of neuroscientific findings on object recognition theories.

Main Methods:

• Review of empirical research and theoretical models in object recognition.
• Analysis of psychophysical evidence and neurophysiological studies.
• Examination of neuroimaging results, including fMRI data.

Main Results:

• Structural-description models dominated early research, followed by view-based accounts challenging their assumptions.
• Neurophysiological studies of the high-level visual cortex offered constraints on object representation.
• fMRI studies revealed category selectivity, further constraining computational models.

Conclusions:

• Significant progress has been made in understanding object recognition through diverse methodologies.
• The field faces the challenge of developing a unified theory that integrates structural, view-based, and neuroscientific findings.
• Further research is needed to explain the complex mechanisms of human object perception and recognition.