Research map / Detection and segmentation / CNN features
DetectoRS: Detecting Objects With Recursive Feature Pyramid and Switchable Atrous Convolution
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 RFP and SAC to HTC | HTC with ResNet-50 · COCO val2017 | box AP: 42.0 -> 49.0 | Helped |
| Add RFP to HTC (ablation) | HTC with ResNet-50 · COCO val2017 | box AP: 42.0 -> 46.2 | Helped |
| Add SAC to HTC (ablation) | HTC with ResNet-50 · COCO val2017 | box AP: 42.0 -> 46.3 | Helped |
| Share backbone stage weights across RFP steps (ablation) | RFP · COCO val2017 | box AP: 46.2 -> 45.4 | Hurt |
| Add the ASPP connecting module (ablation) | RFP · COCO val2017 | box AP: 45.7 -> 46.2 | Helped |
| Add the fusion module (ablation) | RFP · COCO val2017 | box AP: 45.9 -> 46.2 | Helped |
| Increase unrolled RFP steps from 2 to 3 (ablation) | RFP · COCO val2017 | box AP: 46.2 -> 47.5 | Helped |
| Add deformable convolution to SAC (ablation) | SAC · COCO val2017 | box AP: 45.3 -> 46.3 | Helped |
| Add global context modules (ablation) | SAC without deformable convolution · COCO val2017 | box AP: 44.3 -> 45.3 | Helped |
| Use weight locking for the atrous convolution weights (ablation) | SAC without deformable convolution · COCO val2017 | box AP: 44.7 -> 45.3 | Helped |
| Use complementary switches instead of two independent switches (ablation) | SAC without deformable convolution · COCO val2017 | box AP: 45.1 -> 45.3 | Helped |
| Allow the trainable weight difference Δw instead of fixing it to zero (ablation) | SAC · COCO | box AP: +0.1% AP | Helped |
Closest papers by meaning
- Feature Pyramid Networks for Object Detection (CVPR 2017 · similarity 0.90)
- M2Det: A Single-Shot Object Detector Based on Multi-Level Feature Pyramid Network (AAAI 2019 · similarity 0.89)
- Cascade R-CNN: Delving Into High Quality Object Detection (CVPR 2018 · similarity 0.87)
- Seeing without Looking: Contextual Rescoring of Object Detections for AP Maximization (CVPR 2020 · similarity 0.87)
- R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016 · similarity 0.86)
- GiraffeDet: A Heavy-Neck Paradigm for Object Detection (ICLR 2022 · 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.