Research map / Detection and segmentation / CNN features
IQDet: Instance-Wise Quality Distribution Sampling 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 |
|---|---|---|---|
| Use IQDet instead of PAA | ResNet-101-FPN detector · COCO test-dev | AP: 44.6 -> 45.1 | Helped |
| Use IQDet instead of PAA | ResNeXt-64x4d-101 detector · COCO test-dev | AP: 46.3 -> 47.0 | Helped |
| Use IQDet instead of PAA | ResNeXt-64x4d-101-DCN detector · COCO test-dev | AP: 48.6 -> 49.0 | Helped |
| Use IQDet instead of PAA with multi-scale testing | ResNeXt-64x4d-101-DCN detector · COCO test-dev (multi-scale) | AP: 51.3 -> 51.6 | Helped |
| Add QDE to FCOS, using distribution soft labels while retaining grid sampling (ablation) | FCOS · COCO val | AP: 39.4 -> 40.4 | Helped |
| Add center-region QDS with continuous-coordinate sampling (ablation) | FCOS · COCO val | AP: 38.7 -> 39.6 | Helped |
| Make GMM μ, σ, and p learnable (ablation) | IQ-Distribution · COCO val | AP: 38.6 -> 40.0 -> 40.4 | Helped |
| Preserve spatial information by using RoIAlign instead of RoIAlign pooling (ablation) | IQDet feature extractor · COCO val | AP: 40.1 -> 41.1 | Helped |
| Use RoIAlign pooling size 14×14 instead of 7×7 (ablation) | IQDet feature extractor · COCO val | AP: 41.1 -> 41.1 | No clear effect |
| Use quality-distribution sampling rather than uniform sampling (ablation) | IQDet · COCO val | AP: 40.4 -> 40.8 | Helped |
| Use continuous bilinear-interpolated coordinates rather than quantized locations (ablation) | IQDet quality-distribution sampling · COCO val | AP: 40.8 -> 41.1 | Helped |
| Resample 12 rather than 4, 8, 16, or 20 positive predictions (ablation) | IQDet · COCO val | AP: 40.6; 40.6; 41.1; 40.6; 40.5 | Mixed |
Closest papers by meaning
- A High-Resolution Dataset for Instance Detection with Multi-View Object Capture (NeurIPS 2023 · similarity 0.87)
- EfficientDet: Scalable and Efficient Object Detection (CVPR 2020 · similarity 0.86)
- Sampling Techniques for Large-Scale Object Detection From Sparsely Annotated Objects (CVPR 2019 · similarity 0.86)
- COUNTS: Benchmarking Object Detectors and Multimodal Large Language Models under Distribution Shifts (CVPR 2025 · similarity 0.85)
- InstructDET: Diversifying Referring Object Detection with Generalized Instructions (ICLR 2024 · similarity 0.85)
- Plug and Play Active Learning for Object Detection (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.