AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Bias Loss for Mobile Neural Networks

Lusine Abrahamyan, Valentin Ziatchin, Yiming Chen, Nikos Deligiannis

ICCV 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory · 28 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 bias loss (ablation)MobileNetV3 1.0x · ImageNet valtop-1 accuracy: 75.2 -> 75.7Improved
add skip blocks (ablation)MobileNetV3 1.0x (CE loss) · ImageNet valtop-1 accuracy: 75.2 -> 75.5Improved
add bias loss to SkipblockNet-M (ablation)SkipblockNet-M · ImageNet valtop-1 accuracy: 75.5 -> 76.2Improved
add skip blocks instead of widening the network (ablation)MobileNetV3 · ImageNet valtop-1 accuracy: 75.3 -> 75.5Improved
train with bias loss instead of cross-entropyShuffleNetV2 0.5x · CIFAR-100top-1 accuracy: 69.5 -> 71Improved
train with bias loss instead of cross-entropySqueezeNet · CIFAR-100top-1 accuracy: 69.4 -> 70.4Improved
train with bias loss instead of cross-entropyMobileNetV2 0.75x · CIFAR-100top-1 accuracy: 68 -> 68.6Improved
train with bias loss instead of focal lossDenseNet (k=12) · CIFAR-100top-1 accuracy: 79.5 -> 79.9Improved
replace MobileNetV3 Large with SkipblockNet-Mmobile classifier · ImageNet valtop-1 accuracy: 75.2 -> 76.2Improved
replace MobileNetV2 with SkipblockNet-S as SSD backboneSSD · PASCAL VOC detectionmAP: 73.1 -> 73.6Improved
train with bias loss instead of cross-entropy (ablation)SkipblockNet-M · last conv layer feature varianceaverage variance: 0.09 -> 0.15Improved
use SkipblockNet-M instead of MobileNetV3 1.0x (both CE)feature extractor · last conv layer feature varianceaverage variance: 0.05 -> 0.09Improved

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