Related Experiment Videos
Representation of similarity in three-dimensional object discrimination
1Department of Applied Mathematics and Computer Science, Weizmann Institute of Science, Rehovot, Israel.
Neural Computation
|March 1, 1995
Summary
The brain may represent 3D objects using a low-dimensional feature space. This space might be defined by class prototypes, influencing how we perceive object similarities.
Area of Science:
- Cognitive Psychology
- Neuroscience
- Computer Vision
Background:
- The human visual system operates in a low-dimensional psychological space for simple tasks.
- Theories of 3D shape recognition debate feature-space versus structural or pictorial representations.
Purpose of the Study:
- To test the feature-space hypothesis for 3D object recognition.
- To investigate if perceived similarities reflect the structure of an underlying feature space.
Main Methods:
- Subjects were presented with complex parameterized 3D shapes.
- Multidimensional scaling analyzed subjective similarity data.
- Computational simulations were performed.
Main Results:
- Results supported the feature-space hypothesis for 3D object discrimination.
- Evidence suggests a low-dimensional feature space representation is used.
- This space may be spanned by explicitly encoded class prototypes.
Conclusions:
- 3D object discrimination likely relies on a low-dimensional feature space.
- The representation may involve class prototypes.
- This finding has implications for understanding visual object representation.