Research map / Detection and segmentation / CNN features
Dive Deeper Into Box for Object Detection
Ran Chen, Yong Liu, Mengdan Zhang, Shu Liu, Bei Yu, Yu-Wing Tai
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 box decomposition and recombination (D&R) (ablation) | FCOS baseline · COCO minival (ResNet-50) | AP: 33.6 -> 34.8 | Improved |
| add D&R (ablation) | FCOS baseline · COCO minival (ResNet-50) | AP75: 35.0 -> 36.4 | Improved |
| add semantic consistency module (ablation) | FCOS baseline · COCO minival (ResNet-50) | AP: 33.6 -> 37.2 | Improved |
| add semantic consistency module on top of D&R (ablation) | Baseline + D&R · COCO minival (ResNet-50) | AP: 34.8 -> 38.0 | Improved |
| add D&R on top of semantic consistency (ablation) | Baseline + Consistency · COCO minival (ResNet-50) | AP: 37.2 -> 38.0 | Improved |
| add both modules (ablation) | FCOS baseline · COCO minival (ResNet-50) | AP_L: 43.7 -> 50.4 | Improved |
| replace predefined center sampling (PN) with adaptive consistency filtering (ablation) | FCOS baseline · COCO minival (positive assignment, ResNet-50) | AP: 34.2 -> 35.3 | Improved |
| add center sampling (PN) (ablation) | FCOS baseline · COCO minival (positive assignment, ResNet-50) | AP: 33.6 -> 34.2 | Improved |
| add ignore regions between positive and negative (PNI) (ablation) | FCOS baseline · COCO minival (positive assignment, ResNet-50) | AP: 33.6 -> 33.7 | No clear effect |
| add adaptive mean-threshold selection instead of constant ratio c=0.7 (ablation) | semantic consistency filtering · COCO minival (sampling ratio, ResNet-50) | AP: 35.1 -> 35.3 | Improved |
| vary constant sampling ratio c from 0.4 to 0.7 (ablation) | semantic consistency filtering · COCO minival (sampling ratio, ResNet-50) | AP: 34.6 -> 34.1 -> 34.7 -> 35.1 | Mixed |
| add D&R to IoU loss (ablation) | FCOS box regression · COCO trainval35k training IoU | mean IoU of boxes: 0.34 -> 0.47 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- FCOS — FCOS: Fully Convolutional One-Stage Object Detection (ICCV 2019) · better on 2 of 2 shared results
- Cascade R-CNN — Cascade R-CNN: Delving Into High Quality Object Detection (CVPR 2018) · better on 1 of 1 shared result
- CornerNet — CornerNet: Detecting Objects as Paired Keypoints (ECCV 2018) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Bounding Box Regression With Uncertainty for Accurate Object Detection (CVPR 2019 · similarity 0.89)
- Bounding Box Stability against Feature Dropout Reflects Detector Generalization across Environments (ICLR 2024 · similarity 0.87)
- Theoretically Achieving Continuous Representation of Oriented Bounding Boxes (CVPR 2024 · similarity 0.86)
- Union-over-Intersections: Object Detection beyond Winner-Takes-All (ICLR 2025 · similarity 0.86)
- DiffuBox: Refining 3D Object Detection with Point Diffusion (NeurIPS 2024 · similarity 0.85)
- Improving Single Domain-Generalized Object Detection: A Focus on Diversification and Alignment (CVPR 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.
