Research map / Label-efficient and robust learning / Knowledge distillation
Interpretable Foreground Object Search As Knowledge Distillation
Boren Li, Po-Yu Zhuang, Jian Gu, Mingyang Li, Ping Tan
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 interchangeable-foreground pretraining (knowledge distillation) with early fusion (ablation) | baseline FoS · FoS person benchmark (COCO, 100 queries) | mAP: 43.30 -> 48.98 | Improved |
| replace early fusion with late fusion of background and layout (ablation) | FoS with foreground encoder · FoS person benchmark (COCO, 100 queries) | mAP: 48.98 -> 51.91 | Improved |
| add random zoom augmentation of query background (ablation) | late-fusion FoS · FoS person benchmark (COCO, 100 queries) | mAP: 51.91 -> 53.61 | Improved |
| freeze the background encoder (ablation) | late-fusion FoS with augmentation · FoS person benchmark (COCO, 100 queries) | mAP: 53.61 -> 53.72 | No clear effect |
| jointly fine-tune foreground and query encoders with multi-task loss (ablation) | full method · FoS person benchmark (COCO, 100 queries) | mAP: 53.72 -> 54.48 | No clear effect |
| use the proposed distillation framework instead of baseline | FoS · FoS person benchmark (COCO, 100 queries) | mAP: 43.30 -> 53.72 | Improved |
Closest papers by meaning
- Compositing-aware Image Search (ECCV 2018 · similarity 0.90)
- Interpreting Object-level Foundation Models via Visual Precision Search (CVPR 2025 · similarity 0.86)
- FOCUS: Towards Universal Foreground Segmentation (AAAI 2025 · similarity 0.86)
- Referencing Where to Focus: Improving Visual Grounding with Referential Query (NeurIPS 2024 · similarity 0.85)
- DetKDS: Knowledge Distillation Search for Object Detectors (ICML 2024 · similarity 0.85)
- Omni-Q: Omni-Directional Scene Understanding for Unsupervised Visual Grounding (CVPR 2024 · 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.
