Research map / Detection and segmentation / CNN features
A Dynamic Dual-Processing Object Detection Framework Inspired by the Brain's Recognition Mechanism
Minying Zhang, Tianpeng Bu, Lulu Hu
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 NAS-searched feature fusion between CNN and Transformer encoder streams (ablation) | DDP dual-stream encoder · COCO val2017 | AP (CNN stream): 39.2 -> 41.3 | Improved |
| use random fusion architecture instead of searched (ablation) | DDP dual-stream encoder · COCO val2017 | AP (CNN stream): 39.2 -> 37.6 | Hurt |
| use interleaved fusion instead of searched fusion (ablation) | DDP dual-stream encoder · COCO val2017 | AP (CNN stream): 39.2 -> 39.8 | No clear effect |
| add selective mask to dual decoder (ablation) | DDP dynamic dual-decoder · COCO val2017 | AP: 43.8 -> 44.7 | Improved |
| add joint training after mask learning (ablation) | DDP dynamic dual-decoder · COCO val2017 | AP: 44.7 -> 45.2 | Improved |
| replace plain NMS merge of two decoders with selective mask (ablation) | DDP dual decoder · COCO val2017 | AP: 43.8 -> 44.7 | Improved |
| share one backbone for CNN and Transformer branches (ablation) | DDP · COCO val2017 | AP (CNN stream): 37.7 -> 39.2 | No clear effect |
| apply DDP on DAB-DETR-R50 and YOLOF at the same FLOPs | DAB-DETR-R50 · COCO val2017 | AP: 42.2 -> 45.2 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Dynamic DETR — Dynamic DETR: End-to-End Object Detection With Dynamic Attention (ICCV 2021) · 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
- Dynamic DETR: End-to-End Object Detection With Dynamic Attention (ICCV 2021 · similarity 0.88)
- DynamicDet: A Unified Dynamic Architecture for Object Detection (CVPR 2023 · similarity 0.87)
- SODFormer: Streaming Object Detection With Transformer Using Events and Frames (TPAMI 2023 · similarity 0.86)
- Efficient Decoder-Free Object Detection with Transformers (ECCV 2022 · similarity 0.86)
- TGBFormer: Transformer-GraphFormer Blender Network for Video Object Detection (AAAI 2025 · similarity 0.85)
- Transformer brain encoders explain human high-level visual responses (NeurIPS 2025 · similarity 0.84)
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.
