AI Paper Atlas

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

Get the Best of Both Worlds: Improving Accuracy and Transferability by Grassmann Class Representation

Haoqi Wang, Zhizhong Li, Wayne Zhang

ICCV 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory · 4 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 vector class representation with 8-dim Grassmann class subspacesResNet50-D · ImageNet-1K valtop-1 accuracy: 78.04 -> 79.26Improved
replace linear fc with Grassmann fc across architectures8 architectures · ImageNet-1K valtop-1 accuracy: +0.44 to +1.38Improved
use multi-center representations (SoftTriple, SubCenterArcFace) instead of vector classResNet50-D · ImageNet-1K valtop-1 accuracy: 78.04 -> 75.55 (SoftTriple), 77.10 (SubCenterArcFace)Hurt
increase subspace dimension k from 1 to 32 (ablation)GCR ResNet50-D · linear transfer avgaccuracy (%): 73.64 -> 79.70Improved
increase subspace dimension beyond 16 (ablation)GCR ResNet50-D · ImageNet-1K valtop-1 accuracy: 79.21 (k=16) -> 78.63 (k=32)Hurt
normalize features to constant length (ablation)GCR k=8 · ImageNet-1K valtop-1 accuracy: 78.12 -> 79.26Improved
use Riemannian SGD instead of vanilla SGD (ablation)GCR k=8 · ImageNet-1K valtop-1 accuracy: 78.55 -> 79.26Improved
train backbone jointly instead of fine-tuning only the Grassmann layer (ablation)GCR k=8 · ImageNet-1K valtop-1 accuracy: 78.14 -> 79.44Improved
use CosineSoftmax/LabelSmoothing/Dropout instead of softmax for feature transferResNet50-D · linear transfer avgaccuracy (%): 77.98 -> 73.72 (CosineSoftmax)Hurt
gamma 30 instead of 25 (ablation)GCR k=8 · ImageNet-1K valtop-1 accuracy: 79.26 -> 78.47Hurt
geodesic retraction versus QR retraction (ablation)GCR Riemannian SGD · ImageNet-1K valtop-1 accuracy: 79.26 vs 79.13No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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