Research map / Detection and segmentation / CNN features
Probing Deep Into Temporal Profile Makes the Infrared Small Target Detector Much Better
Ruojing Li, Wei An, Yingqian Wang, Xinyi Ying, Yimian Dai, Longguang Wang, Miao Li, Yulan Guo, Li Liu
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 TPro (temporal probe) instead of 1x1xT convolutions (ablation) | DeepPro · NUDT-MIRSDT (SNR<=3 subset) | probability of detection Pd (x1e-2): 92.44 -> 95.84 | Improved |
| add TPro instead of 1x1xT convolutions (ablation) | DeepPro · NUDT-MIRSDT-HiNo | probability of detection Pd (x1e-2): 53.21 -> 59.17 | Improved |
| add TPro instead of 1x1xT convolutions (ablation) | DeepPro · NUDT-MIRSDT | false alarm rate Fa (x1e-5): 1.96 -> 0.72 | Improved |
| increase input length from T=5 to T=40 (ablation) | DeepPro · NUDT-MIRSDT-HiNo | probability of detection Pd (x1e-2): 23.14 -> 59.17 | Improved |
| increase input length from T=10 to T=20 (ablation) | DeepPro · NUDT-MIRSDT-HiNo | probability of detection Pd (x1e-2): 46.44 -> 57.49 | Improved |
| increase input length from T=40 to T=80 (ablation) | DeepPro · NUDT-MIRSDT-HiNo | probability of detection Pd (x1e-2): 59.17 -> 60.21 | No clear effect |
| use 4 SCorMs instead of 1 (ablation) | DeepPro · NUDT-MIRSDT-HiNo | probability of detection Pd (x1e-2): 55.12 -> 59.17 | Improved |
| increase SCorMs from 4 to 8 (ablation) | DeepPro · NUDT-MIRSDT-HiNo | probability of detection Pd (x1e-2): 59.17 -> 59.17 | No clear effect |
| add the 2nd and 3rd levels to the 1st level (ablation) | DeepPro · NUDT-MIRSDT-HiNo | probability of detection Pd (x1e-2): 53.50 -> 59.17 | Improved |
| add spatial-temporal convolutions and keep one level (DeepPro-Plus) (ablation) | DeepPro · NUDT-MIRSDT-HiNo | probability of detection Pd (x1e-2): 59.17 -> 76.23 | Improved |
| increase STD-ResBlocks from n=2 to n=4 (ablation) | DeepPro-Plus · NUDT-MIRSDT-HiNo | probability of detection Pd (x1e-2): 76.23 -> 72.76 | Hurt |
| use three levels (DeepPro+S) instead of one level (DeepPro-Plus) (ablation) | DeepPro-Plus · NUDT-MIRSDT-HiNo | probability of detection Pd (x1e-2): 76.23 -> 76.52 | No clear effect |
Closest papers by meaning
- Rethinking Evaluation of Infrared Small Target Detection (NeurIPS 2025 · similarity 0.89)
- Rethinking Infrared Small Target Detection: A Foundation Driven Efficient Paradigm (NeurIPS 2026 · similarity 0.86)
- IRPruneDet: Efficient Infrared Small Target Detection via Wavelet Structure-Regularized Soft Channel Pruning (AAAI 2024 · similarity 0.86)
- Infrared Small Target Detection with Scale and Location Sensitivity (CVPR 2024 · similarity 0.85)
- Pinwheel-shaped Convolution and Scale-based Dynamic Loss for Infrared Small Target Detection (AAAI 2025 · similarity 0.85)
- SAIST: Segment Any Infrared Small Target Model Guided by Contrastive Language-Image Pretraining (CVPR 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.
