Research map / Label-efficient and robust learning / Recognition and long-tail
CP-mtML: Coupled Projection Multi-Task Metric Learning for Large Scale Face Retrieval
Binod Bhattarai, Gaurav Sharma, Frederic Jurie
Check 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 an age-matching auxiliary task using coupled projections | identity retrieval with LBP features · LFW, no distractors, K=2 | 1-call@K: 38.1 -> 46.1 | Improved |
| add an expression-matching auxiliary task using coupled projections | identity retrieval with LBP features · LFW, 1M distractors, K=5 | 1-call@K: 37.4 -> 43.3 | Improved |
| add an expression-matching auxiliary task using coupled projections | identity retrieval with CNN features · LFW, no distractors, K=5 | 1-call@K: 85.1 -> 86.5 | Improved |
| combine the common projection with the identity-task projection (ablation) | CP-mtML identity retrieval · LFW, LBP, K=20 | 1-call@K: 51.8 -> 69.5 | Improved |
| add the identity-task projection to the common projection (ablation) | CP-mtML identity retrieval · LFW, LBP, K=20 | 1-call@K: 64.8 -> 69.5 | Improved |
| use only the auxiliary-task projection instead of the combined projections (ablation) | CP-mtML identity retrieval · LFW, LBP, K=20 | 1-call@K: 69.5 -> 13.0 | Hurt |
| increase projection dimension from d=32 to d=64 (ablation) | CP-mtML (expr), identity retrieval with LBP · LFW, 1M distractors | 1-call@10: +13.5 | Improved |
| increase projection dimension from d=64 to d=128 (ablation) | CP-mtML (expr), identity retrieval with LBP · LFW, 1M distractors | 1-call@10: +3.5 | Improved |
| increase projection dimension from d=32 to d=64 (ablation) | CP-mtML (expr), identity retrieval with CNN · LFW, 1M distractors | 1-call@10: +1.9 | Improved |
| increase projection dimension from d=64 to d=128 (ablation) | CP-mtML (expr), identity retrieval with CNN · LFW, 1M distractors | 1-call@10: -2.1 | Hurt |
| use a single metric trained on the union of identity and auxiliary-task pairs instead of task-coupled projections | identity retrieval with LBP · LFW, no distractors, K=2 | 1-call@K: 46.1 -> 21.7 | Hurt |
| use identity-matching auxiliary data with coupled projections | age retrieval with CNN features · MORPH age retrieval | 1-call@K | Hurt |
Closest papers by meaning
- Distribution Matching for Multi-Task Learning of Classification Tasks: A Large-Scale Study on Faces & Beyond (AAAI 2024 · similarity 0.86)
- Look across Elapse: Disentangled Representation Learning and Photorealistic Cross-Age Face Synthesis for Age-Invariant Face Recognition (AAAI 2019 · similarity 0.86)
- Dual-Reference Face Retrieval (AAAI 2018 · similarity 0.86)
- Towards Age-Invariant Face Recognition (TPAMI 2020 · similarity 0.86)
- When Age-Invariant Face Recognition Meets Face Age Synthesis: A Multi-Task Learning Framework (CVPR 2021 · similarity 0.85)
- Vec2Face: Scaling Face Dataset Generation with Loosely Constrained Vectors (ICLR 2025 · 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.
