Research map / Label-efficient and robust learning / Knowledge distillation
Bridging Cross-task Protocol Inconsistency for Distillation in Dense Object Detection
Longrong Yang, Xianpan Zhou, Xuewei Li, Liang Qiao, Zheyang Li, Ziwei Yang, Gaoang Wang, Xi Li
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 binary classification distillation loss (sigmoid-based) to GFocal student (ablation) | GFocal-Res50 student · COCO val2017 | mAP: 40.1 -> 42.0 | Improved |
| add IoU-based localization distillation loss (ablation) | GFocal-Res50 student · COCO val2017 | mAP: 40.1 -> 42.3 | Improved |
| combine both classification and localization distillation losses (ablation) | GFocal-Res50 student · COCO val2017 | mAP: 40.1 -> 43.2 | Improved |
| replace LD classification distillation with binary classification distillation loss (ablation) | GFocal Res101-Res50 · COCO val2017 | mAP: 40.4 -> 42.0 | Improved |
| combine proposed loss with PKD feature distillation | RetinaNet-Res50 · COCO val2017 | mAP: 39.7 -> 40.1 | Improved |
| add proposed loss on top of LD | GFocal-Res18 · COCO val2017 | mAP: 38.6 -> 38.9 | Improved |
| vary classification loss weight alpha1 (0.25 to 3.0) (ablation) | binary classification distillation loss · COCO val2017 | mAP: 41.2 to 42.0 across alpha1 0.25..3.0 | No clear effect |
| apply method as self-distillation (student = teacher) (ablation) | GFocal-Res18 · COCO val2017 | mAP: 35.8 -> 36.2 | Improved |
| apply LD as self-distillation | GFocal-Res18 · COCO val2017 | mAP: 35.8 -> 35.0 | Hurt |
| add inheriting initialization from teacher (ablation) | PKD + Ours on FCOS-Res50 · COCO val2017 | mAP: 40.2 -> 40.7 | Improved |
Closest papers by meaning
- Localization Distillation for Dense Object Detection (CVPR 2022 · similarity 0.92)
- Distilling Image Classifiers in Object Detectors (NeurIPS 2021 · similarity 0.90)
- Prediction-Guided Distillation for Dense Object Detection (ECCV 2022 · similarity 0.90)
- Structured Knowledge Distillation for Accurate and Efficient Object Detection (TPAMI 2023 · similarity 0.90)
- Localization Distillation for Object Detection (TPAMI 2023 · similarity 0.90)
- CrossKD: Cross-Head Knowledge Distillation for Object Detection (CVPR 2024 · similarity 0.88)
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.
