Research map / Video / Video segmentation and tracking
Automated Video Face Labelling for Films and TV Material
Omkar Parkhi, Esa Rahtu, Qiong Cao, Andrew Zisserman
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 |
|---|---|---|---|
| use proposed bag formation around subtitle tracks instead of prior bag strategies | MIL face labelling (CNN) · Buffy | average precision: 0.93 -> 0.96 | Improved |
| add speaker detection to bag formation (ablation) | Ours (CNN) · Buffy | average precision: 0.94 -> 0.96 | Improved |
| add speaker detection to bag formation (ablation) | Ours (FV) · Buffy | average precision: 0.88 -> 0.91 | Improved |
| add speaker detection (bag size shrinks) (ablation) | Our bags · Buffy episode 1 | average positive bag size: 2.1 -> 1.4 average bag size | Improved |
| change initialisation of learning algorithm (random vs other methods) (ablation) | Ours · Buffy episodes 1-5 | average precision: 0.9 vs 0.89 | No clear effect |
| add our learning algorithm and bag formation (with negative bags) to baseline (ablation) | Baseline [2] components · Casablanca | average precision: 0.75 -> 0.85 | Improved |
| add background character classifier to [2] components (ablation) | Baseline [2] components · Casablanca | average precision: 0.75 -> 0.80 | Improved |
| add background character classifier to our learning and bags (ablation) | Ours · Casablanca | average precision: 0.85 -> 0.88 | Improved |
| replace [2] track features with our CNN track features (ablation) | Ours · Casablanca | average precision: 0.88 -> 0.93 | Improved |
| add linear programming track labelling constraint (ablation) | Ours · Casablanca | average precision: 0.93 -> 0.93 | No clear effect |
| add linear programming track labelling constraint (ablation) | Ours · Buffy | average precision: 0.96 -> 0.96 | No clear effect |
| add linear programming track labelling constraint (ablation) | Ours · Big Bang Theory | average precision: 0.95 -> 0.96 | Improved |
Closest papers by meaning
- VideoClusterNet: Self-Supervised and Adaptive Face Clustering for Videos (ECCV 2024 · similarity 0.87)
- End-To-End Face Detection and Cast Grouping in Movies Using Erdos-Renyi Clustering (ICCV 2017 · similarity 0.85)
- Video Face Clustering With Unknown Number of Clusters (ICCV 2019 · similarity 0.84)
- Face Association for Videos Using Conditional Random Fields and Max-Margin Markov Networks (TPAMI 2016 · similarity 0.84)
- Tracking Persons-of-Interest via Unsupervised Representation Adaptation (IJCV 2019 · similarity 0.84)
- The Devil of Face Recognition is in the Noise (ECCV 2018 · similarity 0.83)
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.
