Research map / Detection and segmentation / CNN features
Efficient Mirror Detection via Multi-Level Heterogeneous Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace PMD with HetNet | mirror detection · MSD | MAE: 0.047 -> 0.043 | Helped |
| Replace PMD with HetNet | mirror detection · PMD | IoU: 0.660 -> 0.690 | Helped |
| Use heterogeneous modules, with MIC shallow and RSL deep, rather than RSL shallow and MIC deep (ablation) | HetNet · MSD | MAE: 0.046 -> 0.043 | Helped |
| Use MIC at all stages rather than MIC shallow and RSL deep (ablation) | HetNet · MSD | MAE: 0.043 -> 0.049 | Hurt |
| Use RSL at all stages rather than MIC shallow and RSL deep (ablation) | HetNet · MSD | MAE: 0.043 -> 0.046 | Hurt |
| Add GE to the base network (ablation) | HetNet · MSD | MAE: 0.056 -> 0.050 | Helped |
| Add one ICFE instead of MIC (ablation) | Base + GE · MSD | MAE: 0.050 -> 0.053 | Hurt |
| Add MICs without RSLs (ablation) | Base + GE · MSD | MAE: 0.050 -> 0.049 | Helped |
| Add RSLs without MICs (ablation) | Base + GE · MSD | MAE: 0.050 -> 0.049 | Helped |
| Add both MICs and RSLs instead of either module alone (ablation) | Base + GE · MSD | MAE: 0.049 -> 0.043 | Helped |
| Use two same-orientation ICFEs rather than one ICFE (ablation) | MIC module · MSD | MAE: 0.049 -> 0.048 | Helped |
| Use rotated multi-orientation MIC rather than two same-orientation ICFEs (ablation) | MIC module · MSD | MAE: 0.048 -> 0.043 | Helped |
Closest papers by meaning
- Detect Any Mirrors: Boosting Learning Reliability on Large-Scale Unlabeled Data with an Iterative Data Engine (CVPR 2025 · similarity 0.89)
- Weakly-Supervised Mirror Detection via Scribble Annotations (AAAI 2024 · similarity 0.87)
- ZOOM: Learning Video Mirror Detection with Extremely-Weak Supervision (AAAI 2024 · similarity 0.87)
- Effective Video Mirror Detection with Inconsistent Motion Cues (CVPR 2024 · similarity 0.86)
- UNIP: Rethinking Pre-trained Attention Patterns for Infrared Semantic Segmentation (ICLR 2025 · similarity 0.82)
- Implicit Bias of Mirror Flow in Homogeneous Neural Networks: Sparse and Dense Feature Learning (NeurIPS 2026 · similarity 0.82)
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.