Research map / Detection and segmentation / CNN features
Adaptive Spectrum-Aware Feature Disentangled Network for Small Object Detection
Yang Guo, Zihan Yang, Feifei Kou, Yulan Hu, Ran Zhang, Siyuan Yao
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 Adaptive Spectrum Disentanglement (ASD) (ablation) | SFDNet baseline · AI-TOD test | AP: 24.8 -> 28.6 | Improved |
| add Class-Wise Prototype Distillation (CPD) (ablation) | SFDNet baseline · AI-TOD test | AP: 24.8 -> 27.2 | Improved |
| add CPD on top of ASD (ablation) | SFDNet · AI-TOD test | AP: 28.6 -> 29.0 | Improved |
| use multi-spectrum scan instead of sweep scan (ablation) | SFDNet · AI-TOD test | AP: 26.7 -> 29.0 | Improved |
| use multi-spectrum scan instead of neighbor scan (ablation) | SFDNet · AI-TOD test | AP: 26.2 -> 29.0 | Improved |
| use S3M instead of attention for spectrum aggregation (ablation) | SFDNet · AI-TOD test | AP: 26.6 -> 29.0 | Improved |
| use S3M instead of VMamba (ablation) | SFDNet · AI-TOD test | AP: 26.4 -> 29.0 | Improved |
| keep low-frequency spectrum (ablation) | SFDNet · AI-TOD test | AP_vt: 11.4 -> 14.4 | Improved |
| keep mid-frequency spectrum (ablation) | SFDNet · AI-TOD test | AP_vt: 10.0 -> 14.4 | Improved |
| keep high-frequency spectrum (ablation) | SFDNet · AI-TOD test | AP_vt: 9.2 -> 14.4 | Improved |
| vary DoG k from 1.2 to 1.8 (ablation) | SFDNet · AI-TOD test | mAP: 28.8 to 29.1 | No clear effect |
| vary DoG sigma from 0.5 to 2.0 (ablation) | SFDNet · AI-TOD test | mAP: 28.7 to 29.0 | No clear effect |
Closest papers by meaning
- Instance-Invariant Domain Adaptive Object Detection via Progressive Disentanglement (TPAMI 2021 · similarity 0.88)
- Single-Domain Generalized Object Detection in Urban Scene via Cyclic-Disentangled Self-Distillation (CVPR 2022 · similarity 0.88)
- Domain-Invariant Disentangled Network for Generalizable Object Detection (ICCV 2021 · similarity 0.88)
- ScaleKD: Distilling Scale-Aware Knowledge in Small Object Detector (CVPR 2023 · similarity 0.87)
- Triple Feature Disentanglement for One-Stage Adaptive Object Detection (AAAI 2024 · similarity 0.86)
- SET: Spectral Enhancement for Tiny Object Detection (CVPR 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.
