AI Paper Atlas

Research map / Label-efficient and robust learning / Knowledge distillation

Feature Normalized Knowledge Distillation for Image Classification

Kunran Xu, Lai Rui, Yishi Li, Lin Gu

ECCV 2020 · Label-efficient and robust learning › Knowledge distillation · approach: Knowledge distillation for detectors · 106 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace unified temperature with feature-norm-based sample-specific factor (KD-fn)KD, Resnet20 student · Cifar-100accuracy (%): 80.34 -> 81.19Improved
replace unified temperature with feature-norm factor (KD-fn)KD, Resnet56 student · Cifar-100accuracy (%): 81.82 -> 82.23Improved
replace unified temperature with feature-norm factor (KD-fn)KD, Resnet20 student · Cifar-10accuracy (%): 92.86 -> 92.67No clear effect
replace unified temperature with feature-norm factor (KD-fn)KD, Resnet18 student · CUB-200-2011accuracy (%): 80.37 -> 81.52Improved
replace unified temperature with feature-norm factor (KD-fn)KD, Resnet18 student · Stanford Carsaccuracy (%): 90.74 -> 91.53Improved
replace unified temperature with feature-norm factor (KD-fn)KD, Mobilenetv2 student · CUB-200-2011accuracy (%): 81.48 -> 82.40Improved
replace unified temperature with feature-norm factor (KD-fn)KD, Mobilenetv2 student · Stanford Carsaccuracy (%): 90.49 -> 91.10Improved
use a larger teacher (Resnet152 instead of Resnet50) with KDKD, Mobilenetv2 student · CUB-200-2011accuracy (%): 81.48 -> 81.57No clear effect
use a larger teacher (Resnet152 instead of Resnet50) with KD-fnKD-fn, Mobilenetv2 student · CUB-200-2011accuracy (%): 82.40 -> 82.80Improved
use feature-norm factor in self-distillationself KD, Resnet20 · Cifar-100accuracy (%): 79.48 -> 80.16Improved
use feature-norm factor in self-distillationself KD, Mobilenetv2 · CUB-200-2011accuracy (%): 80.32 -> 80.47Improved
add teacher distillation to hypersphere embedding (KD-fn instead of HE)HE, Resnet20 · Cifar-100accuracy (%): 78.96 -> 81.19Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy