AI Paper Atlas

Research map / Detection and segmentation / CNN features

Facial Landmark Detection with Tweaked Convolutional Neural Networks

Yue Wu, Tal Hassner, KangGeon Kim, Gérard G. Medioni, Prem Natarajan

TPAMI 2017 · Detection and segmentation › CNN features · approach: Pedestrian and crowd detection · 188 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 tweaking: fine-tune final layers per FC5 clustervanilla CNN · validation per-cluster errorerror rate: in nearly all clusters, improved by several percentsImproved
add tweaking with alignment-sensitive augmentationvanilla CNN · AFLWerror rate: cut 8% of the errorImproved
add tweaking with alignment-sensitive augmentationvanilla CNN · AFWerror rate: cut 17% of the errorImproved
increase number of clusters K up to 64 (ablation)TCNN · AFW and AFLWerror rate: improves up to K=64, then overfitsImproved
increase clusters beyond K=64 (ablation)TCNN · AFW and AFLWerror rate: overfitting beyond K=64Hurt
tweak at CL4 instead of FC5 (ablation)TCNN · AFW and AFLWerror rate: comparable but more tweak timeNo clear effect
add oversampling and mirroring augmentation (ablation)TCNN · AFW and AFLWerror rate: proved unsuccessfulHurt
add alignment-sensitive augmentation to single-cluster vanilla CNN (ablation)vanilla CNN · AFW and AFLWerror rate: failed to show improved resultsNo clear effect
use TCNN landmarks for alignment instead of ERTrecognition pipeline · Janus CS2 (TAR at FAR=1e-2)true accept rate: .689 -> .718Improved
use TCNN landmarks for alignment instead of LBFrecognition pipeline · Janus CS2 (TAR at FAR=1e-2)true accept rate: .643 -> .718Improved
use TCNN landmarks for alignment instead of CLNFrecognition pipeline · Janus CS2 (TAR at FAR=1e-1)true accept rate: .897 -> .903Improved
initialize CLNF with TCNN five pointsCLNF · 300Wlandmark detection errorsImproved

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