Research map / Label-efficient and robust learning / Knowledge distillation
General Instance Distillation for Object Detection
Check 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 |
|---|---|---|---|
| Apply GID to RetinaNet-Res50 | RetinaNet-Res50 student · COCO val | mAP: 36.2 -> 39.1; +2.9 | Helped |
| Apply GID to FCOS-Res50 | FCOS-Res50 student · COCO val | mAP: 38.5 -> 42.0; +3.5 | Helped |
| Apply GID to R-CNN-Res50 | R-CNN-Res50 student · COCO val | mAP: 38.3 -> 40.2; +1.9 | Helped |
| Apply GID with a ResNet-101 teacher | RetinaNet-MobileNet student · COCO val | mAP: 31.0 -> 33.5; +2.5 | Helped |
| Apply GID | RetinaNet-Res50 student · COCO val, person class | mAP: 50.4 -> 52.8; +2.4 | Helped |
| Use positive GIs rather than GT instances for distillation (ablation) | RetinaNet-Res50 GID · COCO val | mAP: 38.5 -> 38.8 | Helped |
| Add negative GIs to distillation (ablation) | RetinaNet-Res50 GID · COCO val | mAP: 38.2 -> 39.1 | Helped |
| Add relation-based distillation to feature- and response-based distillation (ablation) | RetinaNet-Res50 GID · COCO val | mAP: 38.7 -> 39.1 | Helped |
| Add feature-based distillation (ablation) | RetinaNet-Res50 GID · COCO val | mAP: 36.2 -> 37.9 | Helped |
| Add relation-based distillation (ablation) | RetinaNet-Res50 GID · COCO val | mAP: 36.2 -> 37.3 | Helped |
| Add response-based distillation (ablation) | RetinaNet-Res50 GID · COCO val | mAP: 36.2 -> 37.9 | Helped |
| Increase selected general instances from none to top 5 (ablation) | RetinaNet-Res50 GID · COCO val | mAP: 36.2 -> 38.9 | Helped |
Closest papers by meaning
- Instance-Conditional Knowledge Distillation for Object Detection (NeurIPS 2021 · similarity 0.91)
- Learning Efficient Object Detection Models with Knowledge Distillation (NeurIPS 2017 · similarity 0.90)
- DFD: Distilling the Feature Disparity Differently for Detectors (ICML 2024 · similarity 0.89)
- DCSF-KD: Dynamic Channel-wise Spatial Feature Knowledge Distillation for Object Detection (AAAI 2025 · similarity 0.88)
- DetKDS: Knowledge Distillation Search for Object Detectors (ICML 2024 · similarity 0.86)
- CrossKD: Cross-Head Knowledge Distillation for Object Detection (CVPR 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.