Research map / Detection and segmentation / CNN features
Fine-Grained Dynamic Head for Object Detection
Read the paperCheck 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 the detector head with the fine-grained dynamic head | RetinaNet with ResNet-50 · COCO val | mAP: 35.8 -> 38.2 | Helped |
| Replace the detector head with the fine-grained dynamic head | FreeAnchor with ResNet-50 · COCO val | mAP: 38.3 -> 39.0 | Helped |
| Replace the detector head with the fine-grained dynamic head | FCOS with ResNet-101 · COCO val | mAP: 41.0 -> 42.8 | Helped |
| Replace the detector head with the fine-grained dynamic head | ATSS with ResNet-101 · COCO val | mAP: 41.2 -> 43.2 | Helped |
| Replace the detector head with the fine-grained dynamic head | EfficientDet-D1 · COCO val | mAP: 38.2 -> 38.8 | Helped |
| Use pixel-wise instead of coarse-grained dynamic routing (ablation) | FCOS-D8@L · COCO val | mAP: 38.6 -> 41.0 | Helped |
| Use pixel-wise instead of coarse-grained dynamic routing (ablation) | FCOS-D8@H · COCO val | mAP: 41.4 -> 42.0 | Helped |
| Use the proposed gate activation instead of Softmax (ablation) | FCOS-D8 · COCO val | mAP: 40.2 -> 42.0 | Helped |
| Use the proposed gate activation instead of Restricted Tanh (ablation) | FCOS-D8 · COCO val | FLOPs_avg: 477.6 -> 446.6 | Helped |
| Set the proposed gate activation tau to 0.5 rather than 1.5 (ablation) | FCOS-D8 · COCO val | mAP: 42.0 -> 41.4 | Hurt |
| Set the proposed gate activation tau to 1.0 rather than 1.5 (ablation) | FCOS-D8 · COCO val | mAP: 42.0 -> 41.5 | Hurt |
| Set the proposed gate activation tau to 2.0 rather than 1.5 (ablation) | FCOS-D8 · COCO val | mAP: 42.0 -> 41.8 | Hurt |
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)
- Dynamic Head: Unifying Object Detection Heads With Attentions (CVPR 2021 · similarity 0.89)
- Scale-Equalizing Pyramid Convolution for Object Detection (CVPR 2020 · similarity 0.87)
- DynamicDet: A Unified Dynamic Architecture for Object Detection (CVPR 2023 · similarity 0.87)
- HS-FPN: High Frequency and Spatial Perception FPN for Tiny Object Detection (AAAI 2025 · 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.