AI Paper Atlas

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

Hierarchical Bilinear Pooling for Fine-Grained Visual Recognition

Chaojian Yu, Xinyi Zhao, Qi Zheng, Peng Zhang, Xinge You

ECCV 2018 · Label-efficient and robust learning › Recognition and long-tail · approach: Correspondence and causal representation learning · 368 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 cross-layer interaction relu5_2 (FBP to CBP-1) (ablation)factorized bilinear pooling · CUB-200-2011accuracy (%): 85.70 -> 86.75Improved
add relu5_1 as second cross-layer pair (CBP-1 to HBP-1) (ablation)CBP · CUB-200-2011accuracy (%): 86.75 -> 86.78No clear effect
use relu5_2*relu5_1 pair instead of relu5_3*relu4_3 (HBP-2 to HBP-3) (ablation)HBP · CUB-200-2011accuracy (%): 86.91 -> 87.15Improved
add intermediate layers via HBP-3 over FBP-1 (ablation)FBP · CUB-200-2011accuracy (%): 85.70 -> 87.15Improved
replace relu5_3*relu5_2 with relu5_3*relu4_3 pair (ablation)CBP · CUB-200-2011accuracy (%): 86.75 -> 86.67No clear effect
replace hypercolumn concatenation with cross-layer bilinear poolingHyperBP · CUB-200-2011accuracy (%): 86.60 -> 86.75Improved
replace hypercolumn concatenation with cross-layer bilinear poolingHyperBP · CUB-200-2011model size: 18.4M -> 10.0MImproved
use HBP instead of CBPCBP · CUB-200-2011model size: 10.0M -> 17.5MHurt
increase projection dimension d from 512 to 8192 (ablation)HBP · CUB-200-2011accuracyImproved
add HBP over FBP baselineFBP · Stanford Carsaccuracy (%): 92.8 -> 93.7Improved
use HBP over MA-CNNMA-CNN · FGVC-Aircraftaccuracy (%): 89.9 -> 90.3Improved
use HBP over MA-CNNMA-CNN · CUB-200-2011accuracy (%): 86.5 -> 87.1Improved

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