Research map / Detection and segmentation / CNN features
Joint Detection and Matching of Feature Points in Multimodal Images
Elad Ben Baruch, Yosi Keller
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 hard negative mining (ablation) | Hybrid CNN (CL) · UVIS-NIR ablation (En et al. setup) | FPR95 (%): 9.90 -> 3.41 | Improved |
| add hard negative mining (ablation) | Hybrid CNN (CE) · UVIS-NIR ablation (En et al. setup) | FPR95 (%): 5.78 -> 3.66 | Improved |
| add auxiliary losses L_S and L_A (3 losses vs 1) (ablation) | Hybrid CNN (CL) · UVIS-NIR ablation (En et al. setup) | FPR95 (%): 4.52 -> 3.41 | Improved |
| add auxiliary losses L_S and L_A (3 losses vs 1) (ablation) | Hybrid CNN (CE) · UVIS-NIR ablation (En et al. setup) | FPR95 (%): 4.72 -> 3.66 | Improved |
| use Hybrid instead of symmetric-only network (ablation) | CL, single loss · UVIS-NIR ablation (En et al. setup) | FPR95 (%): 9.56 -> 4.52 | Improved |
| use Hybrid instead of symmetric-only network (ablation) | CE, single loss · UVIS-NIR ablation (En et al. setup) | FPR95 (%): 4.2 -> 4.72 | Hurt |
| use Hybrid instead of asymmetric-only network (ablation) | CE, single loss · UVIS-NIR ablation (En et al. setup) | FPR95 (%): 5.8 -> 4.72 | Improved |
| use Hybrid instead of asymmetric-only network (ablation) | CL, single loss · UVIS-NIR ablation (En et al. setup) | FPR95 (%): 4.32 -> 4.52 | No clear effect |
| use deeper CNN backbone (D-Hybrid-CL) instead of base CNN (ablation) | Hybrid-CL · VIS-NIR (Aguilera split, Mean over 9 categories) | FPR95 (%): 1.93 -> 1.7 | Improved |
| replace Contrastive loss with Cross-Entropy (ablation) | Hybrid CNN · UVIS-NIR (En et al. split) | FPR95 (%): 3.41 -> 3.66 | No clear effect |
| Hybrid CNN versus TS-Net | Hybrid-CL · CUHK (En et al. split) | FPR95 (%): 2.77 -> 0.1 | Improved |
| Hybrid CNN versus HardNet | Hybrid-CL · VIS-NIR (Aguilera split, Mean over 9 categories) | FPR95 (%): 2.80 -> 1.93 | Improved |
Closest papers by meaning
- Reinforced Feature Points: Optimizing Feature Detection and Description for a High-Level Task (CVPR 2020 · similarity 0.86)
- Image Matching from Handcrafted to Deep Features: A Survey (IJCV 2020 · similarity 0.86)
- Convolutional Hough Matching Networks (CVPR 2021 · similarity 0.84)
- Siamese Network for RGB-D Salient Object Detection and Beyond (TPAMI 2021 · similarity 0.84)
- Joint Learning of Semantic Alignment and Object Landmark Detection (ICCV 2019 · similarity 0.84)
- A Unified Multi-scale Deep Convolutional Neural Network for Fast Object Detection (ECCV 2016 · 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.
