Biologically Inspired Visual Model With Preliminary Cognition and Active Attention Adjustment.

Area of Science:

• Computational Neuroscience
• Artificial Intelligence
• Cognitive Science

Background:

• The Hierarchical Max-Pooling (HMAX) model simulates early visual processing but lacks complex cognitive functions.
• Previous work introduced memory and association into HMAX, but a more elaborate framework is needed.
• The primate visual cortex exhibits sophisticated memory, association, and active adjustment capabilities crucial for cognition.

Purpose of the Study:

• To develop an improved computational model of visual cognition by mimicking advanced structures and functions of the primate visual cortex.
• To integrate new mechanisms for memory formation, association, preliminary cognition, and active adjustment in visual processing.
• To enhance object recognition capabilities, particularly under challenging conditions like occlusion and varying orientations.

Main Methods:

• Utilized deep convolutional neural networks to extract diverse episodic features for object recognition.
• Implemented separated feature clusters and loop discharge mechanisms for fast and robust feature retrieval and association.
• Introduced preliminary cognition for object classification and active adjustment for top-down processing (e.g., occlusion, orientation).

Main Results:

• The enhanced model demonstrated efficient visual recognition on the CAS-PEAL-R1 and AR face databases.
• Achieved significantly lower memory storage requirements compared to traditional computational methods.
• Showcased improved performance in object recognition tasks, particularly with complex visual data.

Conclusions:

• The proposed model offers a more sophisticated simulation of visual cognition by integrating advanced neural mechanisms.
• The approach provides a computationally efficient and effective method for robust object recognition.
• This framework advances the understanding of how memory, association, and active adjustment contribute to visual cognition.