Two-dimensional normalized knowledge distillation leveraging class relations

doi:10.1016/j.jvcir.2025.104557

Two-dimensional normalized knowledge distillation leveraging class relations

2025-08-15

0

PRE

AI

DOI：10.1016/j.jvcir.2025.104557

Original

Original request for help

Save

Abstract

En 中文

• Rectification reduces the gap between students and teachers. • Proposed method combines intra-class correlation with inter-class correlation. • Proposed method has excellent training efficiency. • Three image datasets verify the effectiveness of the proposed method.

Keywords:

Rectification

intra-class correlation

inter-class correlation

training efficiency

image datasets

Two-dimensional normalized knowledge distillation leveraging class relations

Abstract

Journal

Organization