Research map / 3D vision / Human pose estimation
Disentangling 3D Pose in a Dendritic CNN for Unconstrained 2D Face Alignment
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Condition landmark estimation on 3D pose (ablation) | PCD-CNN · AFLW validation, PIFA protocol | Normalized Error: 3.45 -> 2.85 | Helped |
| Use classification for the second stage instead of regression (ablation) | Two-stage PCD-CNN · AFLW validation, PIFA protocol | Normalized Error: 3.93 -> 3.09 | Helped |
| Use Mask-Softmax instead of Softmax (ablation) | PCD-CNN · AFLW validation, PIFA protocol | Normalized Error: 4.56 -> 2.85 | Helped |
| Add offline hard sample mining (ablation) | PCD-CNN · AFLW validation, PIFA protocol | Normalized Error: 2.85 -> 2.49 | Helped |
| Increase the number of deconvolution filters (ablation) | PCD-CNN with hard mining · AFLW validation, PIFA protocol | Normalized Error: 2.49 -> 2.40 | Helped |
| Condition the auxiliary localization network on pose (ablation) | PCD-CNN with auxiliary network · AFLW test, PIFA protocol | NME: 2.99 -> 2.49 | Helped |
| Add flipped-image test augmentation and average the outputs | PCD-CNN(C+C) · AFLW-PIFA test | NME: 2.49 -> 2.40 | Helped |
| Add flipped-image test augmentation and average the outputs | PCD-CNN(C+C) · AFW test | NME: 2.52 -> 2.36 | Helped |
| Add flipped-image test augmentation and average the outputs | PCD-CNN · 300W challenge | NME: 7.62 -> 7.17 | Helped |
| Add flipped-image test augmentation and average the outputs | PCD-CNN · COFW test | NME: 6.02 -> 5.77 | Helped |
| Increase the number of deconvolution filters (ablation) | PCD-CNN · AFLW, AFW, and other evaluated datasets | NME | Helped |
| Use the larger deconvolution-filter configuration (ablation) | PCD-CNN · Inference | Frames per second: 20FPS -> 11FPS | Hurt |
Closest papers by meaning
- Learning to Detect 3D Facial Landmarks via Heatmap Regression with Graph Convolutional Network (AAAI 2022 · similarity 0.90)
- PropagationNet: Propagate Points to Curve to Learn Structure Information (CVPR 2020 · similarity 0.87)
- Coarse-To-Fine Volumetric Prediction for Single-Image 3D Human Pose (CVPR 2017 · similarity 0.86)
- Neural Localizer Fields for Continuous 3D Human Pose and Shape Estimation (NeurIPS 2024 · similarity 0.86)
- Unsupervised Discovery of Facial Landmarks and Head Pose (CVPR 2025 · similarity 0.86)
- Wing Loss for Robust Facial Landmark Localisation With Convolutional Neural Networks (CVPR 2018 · similarity 0.86)
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.