Research map / Detection and segmentation / CNN features
Grid R-CNN
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 |
|---|---|---|---|
| Replace box-offset regression with grid-guided localization | Faster R-CNN with ResNet-50 FPN · COCO minival | AP: 37.4 -> 39.6 | Helped |
| Replace box-offset regression with grid-guided localization | Faster R-CNN with ResNet-50 FPN · COCO, IoU=0.9 | AP: 9.6 -> 19.6 | Helped |
| Add supervised grid points beyond two corners (ablation) | Grid R-CNN without feature fusion · COCO minival | AP: 38.3 -> 38.5 -> 38.9 | Helped |
| Add grid-point feature fusion (ablation) | Grid R-CNN · COCO minival | AP: 38.9 -> 39.6 | Helped |
| Use second-order instead of first-order feature fusion (ablation) | Grid R-CNN · COCO minival | AP: 39.2 -> 39.6 | Helped |
| Use second-order instead of first-order feature fusion (ablation) | Grid R-CNN · COCO minival | AP.75: 41.9 -> 42.4 | Helped |
| Enlarge the proposal area directly (ablation) | Grid R-CNN · COCO minival | AP: 37.7 -> 37.7 | No clear effect |
| Enlarge the proposal area directly (ablation) | Grid R-CNN · COCO minival | APsmall: 22.1 -> 20.8 | Hurt |
| Use extended region mapping instead of original proposal mapping (ablation) | Grid R-CNN · COCO minival | AP: 37.7 -> 38.9 | Helped |
| Use extended region mapping instead of original proposal mapping (ablation) | Grid R-CNN · COCO minival | APlarge: 48.0 -> 51.4 | Helped |
| Use grid-guided localization instead of regression | ResNet-50 FPN detector · COCO test-dev | AP: 39.8 -> 41.5 | Helped |
| Use rectangular grid-point localization | Grid R-CNN · COCO test-dev categories | category AP gain | Mixed |
Closest papers by meaning
- G-CNN: An Iterative Grid Based Object Detector (CVPR 2016 · similarity 0.88)
- R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016 · similarity 0.86)
- Multi-Scale Location-Aware Kernel Representation for Object Detection (CVPR 2018 · similarity 0.86)
- Heterogeneous Grid Convolution for Adaptive, Efficient, and Controllable Computation (CVPR 2021 · similarity 0.86)
- LiDAR R-CNN: An Efficient and Universal 3D Object Detector (CVPR 2021 · similarity 0.85)
- Recognize Any Regions (NeurIPS 2024 · similarity 0.83)
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.