AI Paper Atlas

Research map / Detection and segmentation / CNN features

Fine-Grained Species Recognition With Privileged Pooling: Better Sample Efficiency Through Supervised Attention

Andrés C. Rodríguez, Stefano D’Aronco, Konrad Schindler B, Jan Dirk Wegner

TPAMI 2023 · Detection and segmentation › CNN features · approach: Local features and context in CNNs · 7 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 keypoint supervision to attention maps (ablation)AvgPool PrPool · CUB200top-1 accuracy: 84.3 -> 88.1Improved
add keypoint supervision to attention maps (ablation)CovPool PrPool · CUB200top-1 accuracy: 87.1 -> 89.0Improved
add keypoint supervision to attention maps (ablation)AvgPool PrPool · iBirds2017top-1 accuracy: 60.3 -> 65.2Improved
add keypoint supervision to attention maps (ablation)CovPool PrPool · iBirds2017top-1 accuracy: 62.3 -> 67.0Improved
add complementary-map regulariser alone (no keypoint supervision) (ablation)AvgPool attention · CUB200top-1 accuracy: 84.3 -> 87.0Improved
add attention-cropped re-feeding at test time (ablation)AvgPrPool · CUB200top-1 accuracy: 87.7 -> 88.1Improved
add attention-cropped re-feeding at test time (ablation)WSDAN · iBirds2017top-1 accuracy: 54.9 -> 62.4Improved
add attention-cropped re-feeding at test time (ablation)AvgPrPool · CCT20+ Cis locationsoverall accuracy: 81.0 -> 80.3Hurt
add attention-cropped re-feeding at test time (ablation)CovPrPool · CCT20+ Cis locationsoverall accuracy: 76.7 -> 68.8Hurt
increase attention maps from 15 to 32 (add 17 complementary) (ablation)CovPrPool · CUB200top-1 accuracy: 85.6 -> 89.2Improved
increase attention maps from 32 to 64 (ablation)CovPrPool · CUB200top-1 accuracy: 89.2 -> 88.9Hurt
use covariance pooling instead of average pooling with privileged attentionPrPool · CUB200top-1 accuracy: 87.7 -> 89.2Improved

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