AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Resolution Switchable Networks for Runtime Efficient Image Recognition

Yikai Wang, Fuchun Sun, Duo Li, Anbang Yao

ECCV 2020 · Architectures and efficiency › RNNs and normalization · approach: Efficient network compression and training · 33 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
add multi-resolution ensemble distillation (MRED) to parallel training (ablation)RS-Net · ImageNet val 96x96 (ResNet18)top-1 accuracy: 61.5 -> 64.1Improved
parallel training of five resolutions with shared weights and private BN (no MRED)ResNet18 · ImageNet val 96x96top-1 accuracy: 62.6 -> 61.5Hurt
parallel training of five resolutions (no MRED)ResNet18 · ImageNet val 224x224top-1 accuracy: 71.0 -> 73.0Improved
train one RS-Net instead of an individual model at 224 only (I-224)ResNet18 · ImageNet val 96x96top-1 accuracy: 42.5 -> 64.1Improved
add high-to-low resolution distillation paths (full over vanilla MRED) (ablation)MRED ResNet18 · ImageNet val 96x96top-1 accuracy: 63.7 -> 64.1Improved
add distillation from p1..p(S-1) paths in addition to p0 (ablation)MRED ResNet18 · ImageNet val 224x224top-1 accuracy: 72.2 -> 73.1Improved
add distillation from ensemble p0 (vs only p1..p(S-1)) (ablation)MRED ResNet18 · ImageNet val 224x224top-1 accuracy: 72.3 -> 73.1Improved
use single-crop instead of multi-crop across resolutions (ablation)RS-Net ResNet18 · ImageNet val 96x96top-1 accuracy: 62.8 -> 64.1Improved
use multiple resolutions instead of identical resolutions for ensemble distillation (ablation)RS-Net ResNet18 · ImageNet val 224x224top-1 accuracy: 72.1 -> 73.1Improved
train with area ratio [0.3,1.0] instead of [0.08,1.0] to reduce train-test discrepancy (ablation)parallel training ResNet18 · ImageNet valtop-1 accuracy: interaction effects weakenedMixed
adjust resolution instead of width at comparable FLOPsMobileNetV2 I-Nets · ImageNet val 96x96 vs 0.35x widthtop-1 accuracy: 60.1 -> 62.6Improved
apply RS-Net training to quantized ResNet50 (2/2 bits)LQ-Nets ResNet50 · ImageNet val 224x224top-1 accuracy: 72.2 -> 74.0Improved

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