Research map / Label-efficient and robust learning / Recognition and long-tail
Merge or Not? Learning to Group Faces via Imitation Learning
Yue He, Kaidi Cao, Cheng Li, Chen Loy
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 |
|---|---|---|---|
| add imitation-learned sequential merge decisions to hierarchical clustering | HC · LFW-Album | F1 (%): 76.6 -> 91.1 | Improved |
| add imitation-learned sequential merge decisions to hierarchical clustering | HC · ACCIO-1 | F1 (%): 28.9 -> 84.3 | Improved |
| add imitation-learned sequential merge decisions to graph degree linkage | GDL · GFW | F1 (%): 55.9 -> 74.5 | Improved |
| add face quality feature (ablation) | IL-HC · LFW-Album | F1 (%): 89.5 -> 91.1 | Improved |
| add face quality feature (ablation) | IL-HC · ACCIO-1 | F1 (%): 65.0 -> 84.3 | Improved |
| add face quality feature (ablation) | IL-HC · GFW | F1 (%): 48.4 -> 67.3 | Improved |
| add long-term operation-cost reward (ablation) | IL-HC · LFW-Album | F1 (%): 90.7 -> 91.1 | Improved |
| add long-term operation-cost reward (ablation) | IL-HC · GFW | F1 (%): 62.6 -> 67.3 | Improved |
| use the learned short-term reward instead of a naive ±1 ground-truth reward (ablation) | IL-HC · LFW-Album | F1 (%): 73.0 -> 91.1 | Improved |
| use the learned short-term reward instead of a naive ±1 ground-truth reward (ablation) | IL-HC · GFW | F1 (%): 17.1 -> 67.3 | Improved |
| use an HC-based recommender rather than a random pair recommender (ablation) | imitation-learning face grouping · GFW | F1 (%): 61.9% (random recommender) | Improved |
| choose candidate pairs exhaustively using the action-value function rather than using a clustering recommender (ablation) | imitation-learning face grouping · GFW | F1 (%): 82.7% | Improved |
Closest papers by meaning
- End-To-End Face Detection and Cast Grouping in Movies Using Erdos-Renyi Clustering (ICCV 2017 · similarity 0.85)
- GroupFace: Learning Latent Groups and Constructing Group-Based Representations for Face Recognition (CVPR 2020 · similarity 0.85)
- Learning to Cluster Faces via Confidence and Connectivity Estimation (CVPR 2020 · similarity 0.85)
- Linkage Based Face Clustering via Graph Convolution Network (CVPR 2019 · similarity 0.85)
- Learning Group Actions on Latent Representations (NeurIPS 2024 · similarity 0.83)
- Perceptual Group Tokenizer: Building Perception with Iterative Grouping (ICLR 2024 · 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.
