Research map / Detection and segmentation / Open-set and few-shot detection
Query-Based Hard-Image Retrieval for Object Detection at Test Time
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 |
|---|---|---|---|
| Estimate query-specific hardness using Bernoulli score sampling instead of query-agnostic uncertainty | RetinaNet on COCO val · COCO val, PixelAdj(fp) | nDCG: 0.66 -> 0.86 | Helped |
| Estimate query-specific hardness using Bernoulli score sampling instead of query-agnostic uncertainty | Faster-RCNN on COCO val · COCO val, OccAware(fp) | nDCG: 0.85 -> 0.94 | Helped |
| Estimate query-specific hardness using Bernoulli score sampling instead of query-agnostic uncertainty | Mask-RCNN on nuImages val · nuImages val, PixelAdj(fp) | nDCG: 0.55 -> 0.86 | Helped |
| Estimate query-specific hardness using Bernoulli score sampling instead of query-agnostic uncertainty | Cascade Mask-RCNN on nuImages val · nuImages val, PixelAdj(fp) | nDCG: 0.49 -> 0.83 | Helped |
| Use score sampling rather than the strongest uncertainty baseline | RetinaNet on COCO val · COCO val, Total(fn) | nDCG: 0.92 -> 0.92 | No clear effect |
| Use score sampling rather than the strongest uncertainty baseline | Faster-RCNN on COCO val · COCO val, Total(fp) | nDCG: 0.90 -> 0.91 | No clear effect |
| Use score sampling rather than the strongest uncertainty baseline | Mask-RCNN on nuImages val · nuImages val, Total(fn) | nDCG: 0.89 -> 0.89 | No clear effect |
| Use score sampling rather than the strongest uncertainty baseline | Cascade Mask-RCNN on nuImages val · nuImages val, Total(fn) | nDCG: 0.89 -> 0.89 | No clear effect |
| Use 10 Monte Carlo samples for score sampling (ablation) | Score sampling on COCO and nuImages detectors · Ranking evaluation | ranking evaluation metric | Helped |
| Rely on detection scores as Bernoulli probabilities despite detector miscalibration | Score sampling on the evaluated detectors · COCO and nuImages | hard-image ranking | Helped |
Closest papers by meaning
- Large Scale Hard Sample Mining With Monte Carlo Tree Search (CVPR 2016 · similarity 0.87)
- How Hard Can It Be? Estimating the Difficulty of Visual Search in an Image (CVPR 2016 · similarity 0.87)
- QueryDet: Cascaded Sparse Query for Accelerating High-Resolution Small Object Detection (CVPR 2022 · similarity 0.86)
- SearchAD: Large-Scale Rare Image Retrieval Dataset for Autonomous Driving (CVPR 2026 · similarity 0.86)
- Search and Detect: Training-Free Long Tail Object Detection via Web-Image Retrieval (CVPR 2025 · similarity 0.85)
- DetKDS: Knowledge Distillation Search for Object Detectors (ICML 2024 · similarity 0.85)
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.