Research map / Detection and segmentation / CNN features
Synergy Between Face Alignment and Tracking via Discriminative Global Consensus Optimization
Muhammad Haris Khan, John McDonagh, Georgios Tzimiropoulos
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 changed | Where | Measured | Effect |
|---|---|---|---|
| combine detection and tracking via ADMM global consensus instead of detection alone | Detection-conv · 300-VW category C | AUC@0.08 (%): about 6-7% over the detector across the pt-pt-error spectrum | Improved |
| combine detection and tracking via consensus instead of tracking alone | Tracking-conv · 300-VW category C | AUC@0.08 (%): 55.94 -> 58.58 (detector); DGCM higher in most videos | Improved |
| use consensus on video #74 | Detection-conv · 300-VW category C, video #74 | AUC@0.08 (%): 47.43 -> 51.77 | Improved |
| use consensus on video #86 | Detection-conv · 300-VW category C, video #86 | AUC@0.08 (%): 62.10 -> 66.13 | Improved |
| use consensus on video #99 | Detection-conv · 300-VW category C, video #99 | AUC@0.08 (%): 67.88 -> 71.44 | Improved |
| use consensus on video #113 | Detection-conv · 300-VW category C, video #113 | AUC@0.08 (%): 56.00 -> 58.43 | Improved |
| use consensus on video #95 where both detector and tracker work poorly | Detection-conv · 300-VW category C, video #95 | AUC@0.08 (%): 46.96 -> 44.48 | Hurt |
| use consensus on video #100 | Detection-conv · 300-VW category C, video #100 | AUC@0.08 (%): 36.98 -> 35.01 | Hurt |
| use consensus on video #111 | Detection-conv · 300-VW category C, video #111 | AUC@0.08 (%): 63.80 -> 61.38 | Hurt |
| use VGG-16 conv-3 features instead of SIFT | DGCM · 300-VW category C | fitting accuracy | Improved |
| use more than one detection and one tracking initialization in DGCM | DGCM · 300-VW | fitting accuracy | No clear effect |
| use DGCM with 5 initializations instead of PO-CR detector with 5 initializations under high initialization noise | PO-CR detector · 300-W test set and cats dataset | fitting accuracy | Improved |
Closest papers by meaning
- A Functional Regression Approach to Facial Landmark Tracking (TPAMI 2017 · similarity 0.88)
- Cascaded Continuous Regression for Real-Time Incremental Face Tracking (ECCV 2016 · similarity 0.87)
- Face Landmark Fitting via Optimized Part Mixtures and Cascaded Deformable Model (TPAMI 2016 · similarity 0.87)
- Brute-Force Facial Landmark Analysis With a 140,000-Way Classifier (AAAI 2018 · similarity 0.86)
- Supervision-by-Registration: An Unsupervised Approach to Improve the Precision of Facial Landmark Detectors (CVPR 2018 · similarity 0.86)
- 3D Face Tracking from 2D Video through Iterative Dense UV to Image Flow (CVPR 2024 · similarity 0.85)
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.
