A spectral (Fourier) analysis of ViT features guides layer selection and feature alignment for knowledge distillation, yielding ImageNet top-1 gains of +5.2% for DeiT-Tiny and +1.4% for Swin-Tiny.
Dearkd: data-efficient early knowledge distillation for vision transformers
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SpectralKD: A Unified Framework for Interpreting and Distilling Vision Transformers via Spectral Analysis
A spectral (Fourier) analysis of ViT features guides layer selection and feature alignment for knowledge distillation, yielding ImageNet top-1 gains of +5.2% for DeiT-Tiny and +1.4% for Swin-Tiny.