AI Paper Atlas

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

GistNet: A Geometric Structure Transfer Network for Long-Tailed Recognition

Bo Liu, Haoxiang Li, Hao Kang, Gang Hua, Nuno Vasconcelos

ICCV 2021 · Label-efficient and robust learning › Recognition and long-tail · approach: Decoupled and calibrated long-tail training · 57 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 cosine classifier and class-balanced sampling (ablation)Plain Model · ImageNet-LT val (overall)accuracy: 25.1 -> 37.6Improved
add class structure parameters (ablation)COS+CB · ImageNet-LT val (overall)accuracy: 37.6 -> 39.5Improved
add class structure parameters (ablation)COS+CB · ImageNet-LT val (few-shot)accuracy: 14.7 -> 14.5No clear effect
add GIST training (ablation)COS+CS+CB · ImageNet-LT val (overall)accuracy: 39.5 -> 43.5Improved
add GIST training (ablation)COS+CS+CB · ImageNet-LT val (few-shot)accuracy: 14.5 -> 21.4Improved
add class structure parameters (ablation)COS+GIST · ImageNet-LT val (overall)accuracy: 40.2 -> 43.5Improved
use separate centers for random-sampling loss (v_k) rather than shared w_k (ablation)GistNet · ImageNet-LT val (few-shot)accuracy: 14.8 -> 21.4Improved
use separate centers for random-sampling loss (v_k) rather than shared w_k (ablation)GistNet · ImageNet-LT val (many-shot)accuracy: 58.2 -> 54.8Hurt
replace addition g with rotation g (ablation)GistNet · ImageNet-LT val (overall)accuracy: 43.5 -> 43.6No clear effect
replace addition g with MLP g (ablation)GistNet · ImageNet-LT val (overall)accuracy: 43.5 -> 43.4No clear effect
increase structure parameters beyond m=4 (to 8) (ablation)GistNet · ImageNet-LT val (overall)accuracyNo clear effect
increase structure parameters to m=8 (ablation)GistNet · ImageNet-LT val (few-shot)accuracy: one-point gainImproved

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