PIP-PACA: An Interpretable Image Classification Framework via Prototype-Aware Clustering Attention

Area of Science:

• Computer Vision
• Artificial Intelligence
• Machine Learning

Background:

• Deep neural networks (DNNs) excel in image classification accuracy but lack transparency, limiting their use in critical applications.
• Prototype-based methods offer interpretability by linking inputs to semantic prototypes, but struggle with global context and prototype relationships.
• Existing methods like PIP-Net show promise but require improvements in understanding complex image data.

Purpose of the Study:

• To develop an interpretable image classification model, PIP-PACA, that overcomes limitations in global context modeling and prototype relationship utilization.
• To enhance the reliability and transparency of deep learning models in computer vision tasks.
• To improve the efficiency and effectiveness of prototype-based learning frameworks.

Main Methods:

• Proposed a novel prototype-aware clustering attention (PACA) module for interpretable image classification.
• Implemented learnable cluster centers to project features into a prototype space, capturing global context via bidirectional attention.
• Incorporated a normalization operation during feature extraction to stabilize distributions and improve prototype matching.

Main Results:

• PIP-PACA preserves interpretability while achieving significant improvements in classification accuracy.
• The model demonstrates enhanced sparsity and superior prototype purity compared to existing methods.
• The PACA module reduces computational complexity from quadratic to linear, offering greater efficiency.

Conclusions:

• The proposed PACA module effectively enhances prototype learning by improving global context modeling and prototype relationships.
• PIP-PACA offers a more interpretable, accurate, and computationally efficient solution for image classification.
• The findings highlight the potential of clustering-based attention mechanisms in advancing explainable AI for computer vision.