AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

Contour-Hugging Heatmaps for Landmark Detection

James McCouat, Irina Voiculescu

CVPR 2022 · Detection and segmentation › Open-set and few-shot detection · approach: Industrial anomaly detection · 30 citations (OpenAlex)

Check 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
Train on one-hot heatmaps and apply temperature scaling instead of training on Gaussian heatmaps (ablation)U-Net landmark detector · Cephalometric validation setECE: 0.7Improved
Apply temperature scaling to a model trained on one-hot heatmaps (ablation)U-Net landmark detector · Cephalometric validation setcalibrationImproved
Apply temperature scaling to a model trained on Gaussian heatmaps (ablation)Landmark detector · Cephalometric validation setcalibrationNo clear effect
Use one-hot heatmaps rather than Gaussian heatmaps during training (ablation)Landmark detector · Cephalometric datasetcalibrationImproved
Zero heatmap values below 5% of the maximum before computing ERE (ablation)Expected Radial Error · Cephalometric Test Set 1correlation with true radial error: 0.9 to 0.96Improved
Threshold ERE to flag potentially inaccurate landmarksOurs (with uncertainty) · Cephalometric Test Set 2MRE (mm): 1.46 overall; 1.30 Good; 2.43 ErroneousMixed
Use the one-hot classification formulation with the U-Net instead of Gaussian target heatmapsLandmark detector · Cephalometric Test Set 1MRE (mm): 1.20No clear effect

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