AI Paper Atlas

Research map / Architectures and efficiency / CNN design

Compositional Model Based Fisher Vector Coding for Image Classification

Lingqiao Liu, Peng Wang, Chunhua Shen, Lei Wang, Anton van den Hengel, Chao Wang, Heng Tao Shen

TPAMI 2017 · Architectures and efficiency › CNN design · approach: Filter design and trained-filter analysis

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 GMM Fisher coding with sparse compositional FVC (SCFVC)GMMFVC · Birds-200accuracy: 70.1% -> 77.3%Improved
replace GMM Fisher coding with SCFVCGMMFVC · MIT-67accuracy: 72.6% -> 77.6%Improved
replace GMM Fisher coding with SCFVCGMMFVC · Pascal VOC 2007mAP: 84.6% -> 86.6%Improved
add discriminative/residual decomposition guided by supervised coding (HSCFVC over SCFVC) (ablation)SCFVC · Birds-200accuracy: 77.3% -> 80.8%Improved
add discriminative/residual decomposition (HSCFVC over SCFVC) (ablation)SCFVC · MIT-67accuracy: 77.6% -> 79.5%Improved
add discriminative/residual decomposition (HSCFVC over SCFVC) (ablation)SCFVC · Pascal VOC 2007mAP: 86.6% -> 88.1%Improved
add Fisher vector coding on top of the guiding supervised coding (HSCFVC over SupC) (ablation)SupC · Birds-200accuracy: 69.5% -> 80.8%Improved
add residual part Fisher vector to the discriminative part (ablation)HSCFVC · Birds-200, MIT-67, Pascal-07accuracy / mAPImproved
enable guidance term (lambda > 0) in HSCFVC (ablation)HSCFVC · Birds-200, Pascal-07, MIT-67accuracy / mAPImproved
use region-crop scheme with aspect ratio kept and 3 scalesHSCFVC · Pascal VOC 2007mAP: 88.1% -> 89.8%Improved
increase number of GMM Gaussians to 500 while reducing dimension to 400 (ablation)GMMFVC · MIT-67accuracy: 72.6% -> 73.7%Improved
increase Gaussians to 1000 while reducing dimension to 200 (ablation)GMMFVC · MIT-67accuracy: 72.6% -> 71.1%Hurt

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