Research map / Detection and segmentation / CNN features
ELF: Embedded Localisation of Features in Pre-Trained CNN
Assia Benbihi, Matthieu Geist, Cedric Pradalier
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 |
|---|---|---|---|
| use VGG instead of Xception as the backbone for ELF (ablation) | ELF · Webcam | matching score (%): 35.48 -> 43.73 | Improved |
| use VGG instead of AlexNet as the backbone for ELF (ablation) | ELF · HPatches | matching score (%): 35.21 -> 51.84 | Improved |
| use feature-map gradient instead of Laplacian gradient (ablation) | ELF-VGG · HPatches | matching score (%): 55.25 -> 51.84 | No clear effect |
| use feature-map gradient instead of Laplacian gradient (ablation) | ELF-VGG · Webcam | matching score (%): 22.79 -> 43.73 | Improved |
| use feature-map gradient instead of Sobel gradient (ablation) | ELF-VGG · Webcam | matching score (%): 20.99 -> 43.73 | Improved |
| use feature-map gradient instead of Sobel gradient (ablation) | ELF-VGG · HPatches | matching score (%): 42.11 -> 51.84 | Improved |
| replace SuperPoint descriptor with simple ELF-VGG pool3 descriptor | SuperPoint detector · HPatches | matching score (%): 57.11 -> 53.71 | No clear effect |
| replace LF-Net descriptor with simple ELF-VGG pool3 descriptor | LF-Net detector · HPatches | matching score (%): 34.19 -> 44.19 | Improved |
| replace SIFT detector with ELF-VGG detection (SIFT descriptor) | SIFT · Webcam | matching score (%): 8.30 -> 19.14 | Improved |
| replace ORB detector with ELF-VGG detection (ORB descriptor) | ORB · HPatches | matching score (%): 14.76 -> 36.96 | Improved |
| switch simple descriptor from VGG pool3 to pool4 (ablation) | ELF-VGG · Photo-Tourism sparse | matching score (%): 25 -> 26.4 | Improved |
Closest papers by meaning
- CAM Back Again: Large Kernel CNNs from a Weakly Supervised Object Localization Perspective (CVPR 2024 · similarity 0.85)
- R2D2: Reliable and Repeatable Detector and Descriptor (NeurIPS 2019 · similarity 0.85)
- D2-Net: A Trainable CNN for Joint Description and Detection of Local Features (CVPR 2019 · similarity 0.85)
- Attention Weighted Local Descriptors (TPAMI 2023 · similarity 0.84)
- Reinforced Feature Points: Optimizing Feature Detection and Description for a High-Level Task (CVPR 2020 · similarity 0.84)
- XFeat: Accelerated Features for Lightweight Image Matching (CVPR 2024 · similarity 0.83)
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.
