AI Paper Atlas

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

Improving Self-Supervised Lightweight Model Learning via Hard-Aware Metric Distillation

Hao Liu, Mang Ye

ECCV 2022 · Label-efficient and robust learning › Knowledge distillation · approach: Few-shot and metric distillation · 5 citations (OpenAlex)

Read the paper

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
add teacher-guided weighting factors (ablation)SMD · CIFAR-100 supervised distillation, MobileNetV2Top-1 accuracy: 63.48 -> 70.76Improved
add teacher-guided weighting factors (ablation)SMD · CIFAR-100 supervised distillation, resnet20Top-1 accuracy: 69.05 -> 71.56Improved
add L2 embedding alignment in first epochs (ablation)SMD · CIFAR-100 supervised distillation, MobileNetV2Top-1 accuracy: 65.65 -> 70.76Improved
add L2 embedding alignment in first epochs (ablation)SMD · CIFAR-100 supervised distillation, resnet20Top-1 accuracy: 71.64 vs 71.56No clear effect
add the SSL loss to SMD (ablation)SMD · CIFAR-100 self-supervised distillation, resnet32x4 -> MobileNetV2Top-1 accuracy: 61.89 vs 61.32No clear effect
use metric distillation instead of L2 feature alignmentstudent MobileNetV2 · CIFAR-100 self-supervised distillation, resnet32x4 -> MobileNetV2Top-1 accuracy: 58.43 -> 61.89Improved
distill with SMD from a ResNet-50 teacher instead of plain SimSiamResNet-18 · ImageNet linear evalTop-1 accuracy: 30.5 -> 61.8Improved
use SMD instead of CompR-1q under the same training strategyResNet-18 · ImageNet linear evalTop-1 accuracy: 60.5 -> 61.8Improved
use SMD instead of DisCoEfficientNet-B0 · ImageNet linear evalTop-1 accuracy: 66.5 vs 66.5No clear effect
use SMD instead of WCoRD with no label informationMobileNetV2 (ResNet-50 teacher) · CIFAR-100 supervised distillationTop-1 accuracy: 70.45 -> 70.76Improved
cut-off at zero weighting lets student pass teacherWRN-16-2 (WRN-40-2 teacher) · CIFAR-100 supervised distillationTop-1 accuracy: 75.61 -> 76.39Improved
use a stronger distillation signal than the student baselineMobileNetV2 · CIFAR-100 self-supervised distillationTop-1 accuracy: 35.87 -> 61.89Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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