Research map / Label-efficient and robust learning / Knowledge distillation
Denoising-Enhanced Coarse-to-Fine Infrared Small Target Detection with Attention Prior-Guided Knowledge Distillation
Houzhang Fang, Ruixuan Huang, Qiuhuan Chen, Xiaolin Wang, Yi Chang, Luxin Yan
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 denoising-assisted training (DAT) (ablation) | ECFNet baseline · UAV | AP50: 87.26 -> 92.71 | Improved |
| add attention prior-guided KD (APKD) (ablation) | ECFNet baseline · UAV | AP50: 87.26 -> 89.30 | Improved |
| combine DAT and APKD (ablation) | ECFNet · UAV | AP50: 92.71 -> 95.87 | Improved |
| inject target-like noise into GT mask (full DAT) vs concatenating GT mask (ablation) | RBCN · UAV | CRP: 92.62 -> 94.13 | Improved |
| use 20x20 grid instead of 40x40 (ablation) | RBCN · UAV | AP50: 81.62 -> 95.7 | Improved |
| use 20x20 instead of 10x10 grid (ablation) | RBCN · UAV | AP50: 93.84 -> 95.7 | Improved |
| insert DAT in all three RBCN blocks (ablation) | RBCN · UAV | CRP: 89.38 -> 94.13 | Improved |
| insert DAT only in block 2 (ablation) | RBCN · UAV | CRR: 93.45 -> 91.36 | Hurt |
| use C2fP instead of Swin Transformer in RBCN (ablation) | RBCN · UAV | CRP / FLOPs: CRP 96.22 -> 94.13; FLOPs 34.5 -> 8.9 | Mixed |
| use APKD instead of other KD methods (Plain, DRKD, CrossKD, DCSF) (ablation) | fine-stage student · UAV | AP50: best other 91.15 -> 95.87 | Improved |
| apply other KD methods (DRKD, CrossKD, DCSF) without attention prior (ablation) | fine-stage student · UAV | AP50: 92.71 -> 87.70 (CrossKD) | Hurt |
| change teacher among five models (ablation) | APKD · UAV | AP50: 92.71 -> 92.54 to 95.87 | Improved |
Closest papers by meaning
- Rethinking Infrared Small Target Detection: A Foundation Driven Efficient Paradigm (NeurIPS 2026 · similarity 0.89)
- Rethinking Evaluation of Infrared Small Target Detection (NeurIPS 2025 · similarity 0.89)
- IRPruneDet: Efficient Infrared Small Target Detection via Wavelet Structure-Regularized Soft Channel Pruning (AAAI 2024 · similarity 0.88)
- Pinwheel-shaped Convolution and Scale-based Dynamic Loss for Infrared Small Target Detection (AAAI 2025 · similarity 0.88)
- Semi-supervised Infrared Small Target Detection with Thermodynamic-Inspired Uneven Perturbation and Confidence Adaptation (AAAI 2025 · similarity 0.88)
- SAIST: Segment Any Infrared Small Target Model Guided by Contrastive Language-Image Pretraining (CVPR 2025 · 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.
