Research map / Architectures and efficiency / Neural architecture search
Teacher Guided Neural Architecture Search for Face Recognition
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Search width and depth jointly instead of using the hand-crafted student architecture (ablation) | ResNet student with feature knowledge distillation · MegaFace identification (MS-Celeb-1M-v1c-R) | Rank-1 accuracy (%): 87.69 -> 90.40 | Helped |
| Search width and depth jointly instead of using the hand-crafted student architecture (ablation) | ResNet student with feature knowledge distillation · Trillion-Pairs verification (MS-Celeb-1M-v1c-R) | verification accuracy (%): 39.80 -> 40.84 | Helped |
| Add feature knowledge distillation rather than training with AM-Softmax alone (ablation) | ResNet50 student · MegaFace identification (CASIA-WebFace-R) | Rank-1 accuracy (%): 58.60 -> 90.75 | Helped |
| Use feature knowledge distillation rather than probability knowledge distillation (ablation) | ResNet50 student · MegaFace identification (MS-Celeb-1M-v1c-R) | Rank-1 accuracy (%): 91.85 -> 95.52 | Helped |
| Search width only rather than keeping the hand-crafted width (ablation) | ResNet student · MegaFace identification (CASIA-WebFace-R) | Rank-1 accuracy (%): 79.02 -> 82.35 | Helped |
| Search width only rather than keeping the hand-crafted width (ablation) | ResNet student · MegaFace verification (CASIA-WebFace-R) | verification accuracy (%): 80.73 -> 85.21 | Helped |
| Search depth only rather than keeping the hand-crafted depth (ablation) | ResNet student · MegaFace verification (CASIA-WebFace-R) | verification accuracy (%): 90.96 -> 93.29 | Helped |
| Search depth only rather than keeping the hand-crafted depth (ablation) | ResNet student · MegaFace identification (MS-Celeb-1M-v1c-R) | Rank-1 accuracy (%): 95.52 -> 95.86 | Helped |
| Search width only rather than keeping the hand-crafted width (ablation) | ResNet student · MegaFace identification (MS-Celeb-1M-v1c-R) | Rank-1 accuracy (%): 87.69 -> 89.10 | Helped |
| Search depth only rather than keeping the hand-crafted depth (ablation) | ResNet student · LFW (MS-Celeb-1M-v1c-R) | verification accuracy (%): 99.68 -> 99.71 | No clear effect |
Closest papers by meaning
- Search to Distill: Pearls Are Everywhere but Not the Eyes (CVPR 2020 · similarity 0.92)
- Towards Oracle Knowledge Distillation with Neural Architecture Search (AAAI 2020 · similarity 0.91)
- Block-Wisely Supervised Neural Architecture Search With Knowledge Distillation (CVPR 2020 · similarity 0.88)
- Improved Knowledge Distillation via Teacher Assistant (AAAI 2020 · similarity 0.88)
- DisWOT: Student Architecture Search for Distillation WithOut Training (CVPR 2023 · similarity 0.88)
- ScaleKD: Strong Vision Transformers Could Be Excellent Teachers (NeurIPS 2024 · similarity 0.86)
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.