AI Paper Atlas

Research map / Detection and segmentation / CNN features

Snapshot Distillation: Teacher-Student Optimization in One Generation

Chenglin Yang, Lingxi Xie, Chi Su, Alan L. Yuille

CVPR 2019 · Detection and segmentation › CNN features · approach: Efficient network training and design · 218 citations (OpenAlex)

Check 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
Use snapshot distillation instead of direct baseline trainingDenseNet190 · CIFAR100 testclassification error (%): 16.69 -> 16.06Improved
Use snapshot distillation with T=3 instead of T=2 (ablation)DenseNet190 · CIFAR100 testclassification error (%): 16.06 -> 16.33Hurt
Use snapshot distillation instead of direct baseline trainingResNet20 · CIFAR100 testclassification error (%): 33.57 -> 31.76Improved
Use snapshot distillation instead of snapshot ensembleDenseNet190 · CIFAR100 testclassification error (%): 16.70 -> 16.06Improved
Use snapshot distillation instead of direct baseline trainingResNet101 · ILSVRC2012 testTop-1 error (%): 21.62 -> 21.25Improved
Use snapshot distillation instead of direct baseline trainingResNet101 · ILSVRC2012 testTop-5 error (%): 5.80 -> 5.55Improved
Use snapshot distillation instead of direct baseline trainingResNet152 · ILSVRC2012 testTop-1 error (%): 21.17 -> 20.93Improved
Use snapshot distillation instead of direct baseline trainingResNet152 · ILSVRC2012 testTop-5 error (%): 5.66 -> 5.55Improved
Use snapshot distillation instead of snapshot ensembleResNet152 · ILSVRC2012 testTop-1 error (%): 21.84 -> 20.93Improved
Use snapshot distillation instead of snapshot ensembleResNet101 · ILSVRC2012 testTop-1 error (%): 22.14 -> 21.25Improved
Use a pretrained backbone optimized with snapshot distillation rather than the baselineResNet152 · PascalVOC 2007mAP (%): 73.49 -> 74.93Improved
Use a pretrained backbone optimized with snapshot distillation rather than the baselineResNet152 · PascalVOC 2012mIOU (%): 77.53 -> 77.97Improved

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