AI Paper Atlas

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

Towards Calibrated Model for Long-Tailed Visual Recognition from Prior Perspective

Zhengzhuo Xu, Zenghao Chai, Chun Yuan

NeurIPS 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Decoupled and calibrated long-tail training · 5 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 UniMix mixing to CE (ablation)ERM · CIFAR-10-LT rho=100top-1 accuracy: 70.36 -> 76.47Improved
add UniMix mixing to CE (ablation)ERM · CIFAR-100-LT rho=100top-1 accuracy: 38.32 -> 41.46Improved
replace mixup with UniMix (CE loss) (ablation)mixup · CIFAR-10-LT rho=200top-1 accuracy: 67.73 -> 68.42Improved
replace CE with Bayias-compensated CE (ablation)ERM · CIFAR-10-LT rho=100top-1 accuracy: 70.36 -> 78.70Improved
replace CE with Bayias-compensated CE (ablation)ERM · CIFAR-100-LT rho=100top-1 accuracy: 38.32 -> 43.52Improved
add UniMix to Bayias (ablation)Bayias CE · CIFAR-10-LT rho=100top-1 accuracy: 78.70 -> 82.75Improved
add UniMix to Bayias (ablation)Bayias CE · CIFAR-100-LT rho=200top-1 accuracy: 38.83 -> 42.07Improved
add mixup to LDAM (ablation)LDAM · CIFAR-10-LT rho=100top-1 accuracy: 74.47 -> 73.96Hurt
add mixup to LDAM (ablation)LDAM · CIFAR-100-LT rho=100top-1 accuracy: 41.76 -> 40.22Hurt
add UniMix to LDAM (ablation)LDAM · CIFAR-100-LT rho=100top-1 accuracy: 41.76 -> 41.67No clear effect
add UniMix to LDAM (ablation)LDAM · CIFAR-10-LT rho=100top-1 accuracy: 74.47 -> 75.35Improved
use UniMix+Bayias instead of ERM for calibrationERM · CIFAR-100-LT rho=100ECE: 38.23 -> 23.04Improved

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