Research map / Detection and segmentation / Open-set and few-shot detection
Sampling Techniques for Large-Scale Object Detection From Sparsely Annotated Objects
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 |
|---|---|---|---|
| Use pseudo labels to ignore likely unannotated-object regions | Object detector trained on sparsely annotated COCO · Sparse COCO (α=0.3, 0.5, 0.7) | mmAP: 34.22 -> 35.00; 31.69 -> 32.79; 27.31 -> 29.03 | Helped |
| Use overlap-based soft sampling to downweight negative losses | Object detector trained on sparsely annotated COCO · Sparse COCO (α=0.3, 0.5, 0.7) | mmAP: 34.22 -> 33.98; 31.69 -> 31.39; 27.31 -> 27.30 | Hurt |
| Use part-aware sampling to ignore losses for unverified parts inside subject boxes | Object detector trained on OID · Open Images v4 validation | mAP: 64.49 -> 65.18 | Helped |
| Use pseudo label-guided sampling | Object detector trained on OID · Open Images v4 validation | mAP: 64.49 -> 64.84 | Helped |
| Use part-aware sampling for part categories | Object detector trained on OID · OID part categories, validation | Average AP: 61.1 -> 70.3 | Helped |
| Use part-aware sampling for human part categories | Object detector trained on OID · OID human part categories | Average AP improvement: 22.7 AP | Helped |
| Select pseudo-label thresholds per category using precision rather than a uniform threshold (ablation) | Pseudo label-guided sampling · Sparse COCO | mmAP: 34.76 -> 34.94 | Helped |
| Set the category-wise minimum precision threshold to >0.5 (ablation) | Pseudo label-guided sampling · Sparse COCO | mmAP: 34.94 -> 35.00 | Helped |
| Raise the category-wise minimum precision threshold from >0.5 to >0.7 (ablation) | Pseudo label-guided sampling · Sparse COCO | mmAP: 35.00 -> 34.93 | Hurt |
| Assign proposals overlapping pseudo labels positive instead of ignoring their loss (ablation) | Pseudo label-guided sampling · Sparse COCO | mmAP: 35.00 -> 34.59 | Hurt |
| Add 16-epoch training (ablation) | Part-aware OID detector · OID validation, single-scale model | mAP: +0.6 | Helped |
| Add a context head (ablation) | Part-aware OID detector · OID validation, single-scale model | mAP: +0.2 | Helped |
Closest papers by meaning
- COUNTS: Benchmarking Object Detectors and Multimodal Large Language Models under Distribution Shifts (CVPR 2025 · similarity 0.87)
- Plug and Play Active Learning for Object Detection (CVPR 2024 · similarity 0.87)
- Object Detection Benchmarks are Incomplete: The Role of Label Errors and Annotation Uncertainty (NeurIPS 2026 · similarity 0.87)
- Open-Det: An Efficient Learning Framework for Open-Ended Detection (ICML 2025 · similarity 0.86)
- Open-World Objectness Modeling Unifies Novel Object Detection (CVPR 2025 · similarity 0.86)
- Multi-Part Object Representations via Graph Structures and Co-Part Discovery (NeurIPS 2026 · 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.