Research map / Label-efficient and robust learning / Recognition and long-tail
Kinship Identification through Joint Learning using Kinship Verification Ensembles
Wei Wang, Shaodi You, Theo Gevers
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 joint learning of verification and identification (JLNet vs Multi-class Net) | Multi-class Net · KinfaceWII Mixed Kin-Type set | accuracy: 0.5225 -> 0.5790 | Improved |
| add joint learning (JLNet vs Ensemble Net) | Ensemble Net · KinfaceWI Mixed Kin-Type set | accuracy: 0.4887 -> 0.5993 | Improved |
| add backpropagation of multiple outputs during joint training (ablation) | JLNet · KinfaceWI Mixed Kin-Type set | accuracy: 0.5487 -> 0.5993 | Improved |
| add backpropagation of multiple outputs during joint training (ablation) | JLNet · KinfaceWII Mixed Kin-Type set | macro F1: 0.4648 -> 0.5343 | Improved |
| combine verification and multi-class outputs at test time instead of multiple output only (ablation) | JLNet · KinfaceWI Mixed Kin-Type set | macro F1: 0.5507 -> 0.5506 | No clear effect |
| combine verification and multi-class outputs at test time instead of multiple output only (ablation) | JLNet · KinfaceWI Real-Scenario set | accuracy: 0.6681 -> 0.7439 | Improved |
| add backpropagation of multiple outputs during joint training (ablation) | JLNet · KinfaceWII Real-Scenario set | accuracy: 0.5730 -> 0.7398 | Improved |
| train verification models on Mixed set instead of independent sets (Ensemble Net vs Ensemble Net*) | Ensemble Net · KinfaceWI Mixed Kin-Type set | macro F1: 0.3240 -> 0.4837 | Improved |
| train verification models on Mixed set instead of independent sets | Ensemble Net · KinfaceWI Independent Kin-Type set (verification) | mean accuracy: 0.7021 -> 0.6224 | Hurt |
| JLNet instead of Ensemble Net* for verification on independent set | Ensemble Net* · KinfaceWI Independent Kin-Type set (verification) | mean accuracy: 0.7021 -> 0.6755 | Hurt |
| add backpropagation of multiple outputs during joint training (ablation) | JLNet · KinfaceWII Independent Kin-Type set (verification) | mean accuracy: 0.6190 -> 0.6965 | Improved |
| JLNet instead of Multi-class Net on real-scenario retrieval | Multi-class Net · KinfaceWI Real-Scenario set | F10(all): 0.2947 -> 0.3459 | Improved |
Closest papers by meaning
- Meta-Mining Discriminative Samples for Kinship Verification (CVPR 2021 · similarity 0.90)
- A Unified Approach to Kinship Verification (TPAMI 2020 · similarity 0.90)
- Cross-Generation Kinship Verification with Sparse Discriminative Metric (TPAMI 2018 · similarity 0.88)
- Facial Kinship Verification: A Comprehensive Review and Outlook (IJCV 2022 · similarity 0.86)
- Visual Kinship Recognition of Families in the Wild (TPAMI 2018 · similarity 0.86)
- T-NET: Weakly Supervised Graph Learning for Combatting Human Trafficking (AAAI 2024 · similarity 0.76)
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.
