Research map / Detection and segmentation / CNN features
R(Det)2: Randomized Decision Routing 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 |
|---|---|---|---|
| Add the R(Det)² decision head and randomized routing | Faster R-CNN with ResNet-50 · COCO val | AP: 37.4 -> 40.4 | Helped |
| Add routing masks to the R(Det)² decision head (ablation) | Faster R-CNN with 2fc features · COCO val | AP: 38.8 -> 39.1 | Helped |
| Use task disentanglement in the R(Det)² head (ablation) | Faster R-CNN with residual 4conv1fc features · COCO val | AP: 40.1 -> 40.4 | Helped |
| Use Focal instead of CE classification loss with Smooth-L1 regression (ablation) | Faster R-CNN + R(Det)², ResNet-50 · COCO val | AP: 40.4 -> 40.5 | Helped |
| Use IoU instead of Smooth-L1 bbox regression loss (ablation) | Faster R-CNN + R(Det)², ResNet-50 with Focal classification · COCO val | AP: 40.5 -> 41.0 | Helped |
| Reduce the selective-loss weight to λ=0.001 (ablation) | R(Det)² routing · COCO val | AP: 38.6%; by 1.8% points | Hurt |
| Set λ=0.95, nearly removing the associative-loss contribution (ablation) | R(Det)² routing · COCO val | AP: 0.3~0.4% | Hurt |
| Add R(Det)² routing to the detector | Libra R-CNN · COCO val | AP: +3.1 | Helped |
| Add R(Det)² routing to the detector | Cascade R-CNN · COCO val | AP: +2.2 | Helped |
| Add R(Det)² routing to the detector | Dynamic R-CNN · COCO val | AP: +2.1 | Helped |
| Add randomized-routing training to the existing task-aware head | DoubleHead R-CNN · COCO val | AP: 1.4% | Helped |
| Add randomized-routing training to the detector | RetinaNet · COCO val | AP: 1.8% | Helped |
Closest papers by meaning
- DetNAS: Backbone Search for Object Detection (NeurIPS 2019 · similarity 0.86)
- Learning Detection With Diverse Proposals (CVPR 2017 · similarity 0.86)
- Hybrid Proposal Refiner: Revisiting DETR Series from the Faster R-CNN Perspective (CVPR 2024 · similarity 0.85)
- Order Matters: Competition-Guided Query Ordering for RNN-Based Object Detection (NeurIPS 2026 · similarity 0.84)
- DetKDS: Knowledge Distillation Search for Object Detectors (ICML 2024 · similarity 0.84)
- MS-DETR: Efficient DETR Training with Mixed Supervision (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.