Research map / Detection and segmentation / CNN features
Transformer-Based Context Condensation for Boosting Feature Pyramids in Object Detection
Zhe Chen, Jing Zhang, Yufei Xu, Dacheng Tao
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 TCC refinement instead of 3x3 conv (ablation) | FPN ResNet50 · MS COCO val | AP: 37.4 -> 40.3 | Improved |
| add TCC (also cuts compute) | FPN ResNet50 · MS COCO test-dev | AP: 37.3 -> 40.6 | Improved |
| add local contexts for MFF refinement (ablation) | FPN ResNet50 · MS COCO val | AP: 37.4 -> 38.9 | Improved |
| add globally summarized key contexts to local contexts (no Transformer) (ablation) | FPN ResNet50 · MS COCO val | AP: 38.9 -> 39.3 | Improved |
| add Transformer decoder to condensed contexts (ablation) | FPN ResNet50 · MS COCO val | AP: 39.3 -> 40.3 | Improved |
| refine fusion output instead of no refinement (ablation) | FPN ResNet50 · MS COCO val | AP: 36.9 -> 40.3 | Improved |
| increase key global contexts from 1 to 4 (ablation) | TCC · MS COCO val | AP: 39.8 -> 40.3 | Improved |
| increase key global contexts from 4 to 8 (ablation) | TCC · MS COCO val | AP: 40.3 -> 40.2 | No clear effect |
| add TCC before fusion in addition to after fusion (ablation) | TCC on MFF · MS COCO val | AP: 0.4 increase in AP | Improved |
| add TCC to NasFPN (640 / 1024 input) | NasFPN · MS COCO test-dev | AP: 39.9 and 44.1 to 41.6 and 45.4 | Improved |
| add TCC to FPG (R50 / R101) | FPG · MS COCO test-dev | AP: 42.4 and 43.8 to 44.1 and 45.2 | Improved |
| add TCC to RFP (HTC) | RFP · MS COCO test-dev | AP: 50.7 in AP to 51.5 in AP | Improved |
Closest papers by meaning
- Hierarchical Context Embedding for Region-based Object Detection (ECCV 2020 · similarity 0.90)
- Feature Pyramid Transformer (ECCV 2020 · similarity 0.90)
- Context-Transformer: Tackling Object Confusion for Few-Shot Detection (AAAI 2020 · similarity 0.90)
- Feature Pyramid Networks for Object Detection (CVPR 2017 · similarity 0.89)
- ContextShift: A Controlled Benchmark for Context Dependence in Object Detection (NeurIPS 2026 · similarity 0.88)
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 2024 · similarity 0.87)
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.
