AI Paper Atlas

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

Solving Long-tailed Recognition with Deep Realistic Taxonomic Classifier

Tz-Ying Wu, Pedro Morgado, Pei Wang, Chih-Hui Ho, Nuno Vasconcelos

ECCV 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Decoupled and calibrated long-tail training · 45 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
replace Softmax with Deep-RTCiNaturalist classifier · iNaturalist 2018 (all classes)CPB: 0.66 -> 0.78Improved
replace Softmax with Deep-RTCfew-shot classes on iNaturalist · iNaturalist 2018 (few-shot classes)CPB: 0.62 -> 0.75Improved
replace Softmax with Deep-RTCmany-shot classes on iNaturalist · iNaturalist 2018 (many-shot classes)CPB: 0.76 -> 0.84Improved
use CBLoss instead of SoftmaxiNaturalist classifier · iNaturalist 2018 (many-shot classes)CPB: 0.76 -> 0.61Hurt
use tau-norm instead of SoftmaxiNaturalist classifier · iNaturalist 2018 (many-shot classes)CPB: 0.76 -> 0.71Hurt
replace Softmax with Deep-RTCImageNet-LT classifier · ImageNet-LTCPB: 0.35 -> 0.53Improved
replace flat realistic predictor with Deep-RTC at 20% rejectionCIFAR100-LT · CIFAR100-LT, 20% rejection (all classes)CPB: .433 -> .491Improved
replace flat realistic predictor with Deep-RTC at 5% rejectionAWA2-LT · AWA2-LT, 5% rejection (all classes)CPB: .914 -> .931Improved
replace flat realistic predictor with Deep-RTC on many-shot classes at 5% rejectionCIFAR100-LT · CIFAR100-LT, 5% rejection (many-shot classes)CPB: .779 -> .773No clear effect
add stochastic tree sampling and node-consistency loss to parameter inheritance (Deep-RTC) (ablation)recursive hierarchical classifier · iNaturalist-subCPBImproved
add stochastic tree sampling (global) instead of node-consistency loss (local) (ablation)parameter-inheritance classifier · iNaturalist-subCPBImproved
use bottom-up hierarchical inference (RHC) instead of flat softmax (ablation)flat classifier · iNaturalist-subCPBImproved

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