Research map / 3D vision / Human pose estimation
Dynamic Facial Analysis: From Bayesian Filtering to Recurrent Neural Network
Check 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 |
|---|---|---|---|
| Use recurrent temporal modeling instead of frame-wise estimation (ablation) | Facial landmark localization model · 300-VW Split 1 | AUC: 0.66 -> 0.74 | Helped |
| Use recurrent temporal modeling instead of frame-wise estimation (ablation) | Facial landmark localization model · 300-VW Split 2 | FR (%): 0.14 -> 0.01 | Helped |
| Use recurrent temporal modeling instead of frame-wise estimation (ablation) | Facial landmark localization model · 300-VW Split 3 | AUC: 0.67 -> 0.77 | Helped |
| Replace Kalman filtering with learned recurrent tracking (ablation) | Facial landmark localization model · 300-VW Split 1 | AUC: 0.63 -> 0.74 | Helped |
| Replace particle filtering with learned recurrent tracking (ablation) | Facial landmark localization model · 300-VW Split 2 | AUC: 0.69 -> 0.77 | Helped |
| Replace learned recurrent tracking with Kalman filtering (ablation) | Facial landmark localization model · 300-VW Split 3 | FR (%): 0.04 -> 0.68 | Hurt |
| Fine-tune the SynHead-trained model with BIWI training data | Head pose estimation model · BIWI | pitch error (°): 3.48 -> 1.35 | Helped |
| Fine-tune the SynHead-trained model with BIWI training data | Head pose estimation model · BIWI | yaw error (°): 3.14 -> 1.54 | Helped |
| Fine-tune the SynHead-trained model with BIWI training data | Head pose estimation model · BIWI | roll error (°): 2.60 -> 1.35 | Helped |
| Use the end-to-end RNN rather than frame-wise estimation (ablation) | Head pose estimation model · SynHead | pitch error (°): 1.94 -> 1.55 | Helped |
| Use the end-to-end RNN rather than Kalman filtering (ablation) | Head pose estimation model · BIWI | yaw error (°): 4.15 -> 3.14 | Helped |
Closest papers by meaning
- Recurrence Without Recurrence: Stable Video Landmark Detection With Deep Equilibrium Models (CVPR 2023 · similarity 0.87)
- LSTM Pose Machines (CVPR 2018 · similarity 0.86)
- Continuous Landmark Detection With 3D Queries (CVPR 2023 · similarity 0.86)
- Mnemonic Descent Method: A Recurrent Process Applied for End-To-End Face Alignment (CVPR 2016 · similarity 0.86)
- Unsupervised Discovery of Facial Landmarks and Head Pose (CVPR 2025 · similarity 0.86)
- Trend-Aware Supervision: On Learning Invariance for Semi-supervised Facial Action Unit Intensity Estimation (AAAI 2024 · similarity 0.84)
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.