Research map / Detection and segmentation / CNN features
Bottom-Up Object Detection by Grouping Extreme and Center Points
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 multi-scale testing (ablation) | ExtremeNet · COCO val2017 | AP: 40.3 -> 43.3 | Helped |
| remove center grouping (ablation) | ExtremeNet · COCO val2017 | AP: 40.3 -> 38.2 | Hurt |
| add edge aggregation (ablation) | ExtremeNet · COCO val2017 | AP: 0.7% | Helped |
| add ghost box suppression (ablation) | ExtremeNet · COCO val2017 | AP: 0.3% | Helped |
| replace predicted center heatmap with ground truth (ablation) | ExtremeNet · COCO val2017 | AP: 48.6 | Helped |
| replace predicted extreme point heatmaps with ground truth (ablation) | ExtremeNet · COCO val2017 | AP: 56.3 | Helped |
| replace predicted extreme points and centers with ground truth (ablation) | ExtremeNet · COCO val2017 | AP: 79.8 | Helped |
| replace predicted extreme points, centers, and offsets with ground truth (ablation) | ExtremeNet · COCO val2017 | AP: 86.0 | Helped |
| use associative embedding instead of geometric center grouping (ablation) | ExtremeNet · COCO val2017 | AP: 2.1% AP drop | Hurt |
| form an octagon mask from predicted extreme points instead of filling the bounding box (ablation) | ExtremeNet · COCO val2017 | Mask AP: 12.1 -> 18.9 | Helped |
| guide DEXTR with predicted extreme points (ablation) | ExtremeNet · COCO val2017 | Mask AP: 34.6 | Helped |
Closest papers by meaning
- Deep Extreme Cut: From Extreme Points to Object Segmentation (CVPR 2018 · similarity 0.87)
- Precise Detection in Densely Packed Scenes (CVPR 2019 · similarity 0.87)
- Feature Pyramid Networks for Object Detection (CVPR 2017 · similarity 0.87)
- Don't Even Look Once: Synthesizing Features for Zero-Shot Detection (CVPR 2020 · similarity 0.86)
- R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016 · similarity 0.86)
- Unsupervised Object Detection with Theoretical Guarantees (NeurIPS 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.