AI Paper Atlas

Research map / Detection and segmentation / CNN features

Localizing by Describing: Attribute-Guided Attention Localization for Fine-Grained Recognition

Xiao Liu, Jiang Wang, Shilei Wen, Errui Ding, Yuanqing Lin

AAAI 2017 · Detection and segmentation › CNN features · approach: Local features and context in CNNs · 94 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
Use ground-truth bird bounding boxes during localizationAttribute-guided localizers · CUB-200-2011PCP (average): 66.2 -> 73.4Improved
Use localized attention regions rather than full-image input alone (ablation)Attribute prediction · CUB-200-2011AUC (four-part average): 74.3 -> 73.7Hurt
Combine full-image and localized-attention predictions (ablation)Attribute prediction · CUB-200-2011AUC (four-part average): 74.3 -> 76.6Improved
Add head and breast part features (ablation)Full-image ResNet-50 classifier · CUB-200-2011, without BBAccuracy: 81.7% -> 84.7%Improved
Add head and breast part features (ablation)Full-image ResNet-50 classifier · CUB-200-2011, with BBAccuracy: 82.3% -> 84.9%Improved
Add features from four localized parts (ablation)Full-image ResNet-50 classifier · CUB-200-2011, without BBAccuracy: 81.7% -> 85.1%Improved
Add features from four localized parts (ablation)Full-image ResNet-50 classifier · CUB-200-2011, with BBAccuracy: 82.3% -> 85.3%Improved
Add binary attribute labels (ablation)Full-image ResNet-50 classifier · CUB-200-2011, without BBAccuracy: 81.7% -> 81.7%No clear effect
Add binary attribute labels (ablation)Full-image ResNet-50 classifier · CUB-200-2011, with BBAccuracy: 82.3% -> 82.2%No clear effect
Add visual features from part attribute-prediction models (ablation)Full-image ResNet-50 classifier · CUB-200-2011, without BBAccuracy: 81.7% -> 82.5%Improved
Add visual features from part attribute-prediction models (ablation)Full-image ResNet-50 classifier · CUB-200-2011, with BBAccuracy: 82.3% -> 82.9%Improved
Combine four-part features with attribute-model features (ablation)Full-image ResNet-50 classifier · CUB-200-2011, without BBAccuracyImproved

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