AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

On Learning Discriminative Features from Synthesized Data for Self-Supervised Fine-Grained Visual Recognition

Zihu Wang, Lingqiao Liu, Scott Ricardo Figueroa Weston, Samuel Tian, Peng Li

ECCV 2024 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Invariance-based self-supervised learning · 9 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 generated perturbed-pair contrastive loss on top of MoCo v2MoCo v2 · CUB-200 linear evaluationtop-1 accuracy: 63.98 -> 66.17Improved
add generated perturbed-pair contrastive loss on top of MoCo v2MoCo v2 · Stanford Cars linear evaluationtop-1 accuracy: 62.02 -> 65.60Improved
add generated perturbed-pair contrastive loss on top of SimSiamSimSiam · CUB-200 linear evaluationtop-1 accuracy: 46.75 -> 57.80Improved
add method on label-scarce linear evaluationMoCo v2 · average of three FGVR datasetstop-1 accuracy gain over MoCo v2: +3.31%, +4.27%, +5.51% with 100%, 50%, 20% labelsImproved
add method on non-fine-grained dataMoCo v2 · CIFAR-100top-1 accuracy: 61.79 -> 63.26Improved
add method on non-fine-grained dataMoCo v2 · ImageNet-100top-1 accuracy: 62.45 -> 64.02Improved
add low-variance dimension noise to Grad-CAM-guided noise (ablation)Ours · Stanford Cars linear evaluationtop-1 accuracy: 64.18 -> 65.60Improved
add Grad-CAM-guided noise to low-variance noise (ablation)Ours · FGVC-Aircraft linear evaluationtop-1 accuracy: 53.11 -> 55.28Improved
add reconstruction loss to MoCo v2 (ablation)MoCo v2 · Stanford Cars linear evaluationtop-1 accuracy: less than 1%Improved

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