Research map / Label-efficient and robust learning / Recognition and long-tail
Beyond Cartesian Representations for Local Descriptors
Patrick Ebel, Anastasiia Mishchuk, Kwang Moo Yi, Pascal Fua, Eduard Trulls
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 log-polar instead of cartesian patches at lambda=12 (ablation) | descriptor network · New dataset | FPR95: 0.72 -> 0.67 | Improved |
| enlarge support region to lambda=96 with log-polar patches (ablation) | log-polar descriptor · New dataset | FPR95: 0.67 -> 0.36 | Improved |
| enlarge support region from lambda=12 to 96 with cartesian patches (ablation) | cartesian descriptor · New dataset | FPR95: 0.72 -> 7.03 | Hurt |
| increase lambda beyond 96 for log-polar patches (ablation) | log-polar descriptor · New dataset | FPR95: 0.36 -> 0.36 | No clear effect |
| use log-polar instead of cartesian at lambda=96 on AMOS (ablation) | descriptor network · AMOS patches | rank-1: 0.481 -> 0.651 | Improved |
| enlarge lambda from 12 to 96 with log-polar patches (ablation) | log-polar descriptor · HPatches illumination split | rank-1: 0.729 -> 0.774 | Improved |
| enlarge lambda from 12 to 96 with cartesian patches (ablation) | cartesian descriptor · HPatches viewpoint split | rank-1: 0.828 -> 0.681 | Hurt |
| enlarge lambda from 12 to 96 with log-polar patches (ablation) | log-polar descriptor · HPatches viewpoint split | rank-1: 0.833 -> 0.847 | Mixed |
| enlarge lambda from 16 to 96 with log-polar patches (ablation) | log-polar descriptor · PhotoTourism stereo task | mAP at 15 deg: 0.0420 -> 0.0448 | Improved |
| use log-polar lambda=96 instead of cartesian lambda=16 (ablation) | descriptor network · PhotoTourism multi-view task | mAP at 15 deg: 0.5208 -> 0.5427 | Improved |
| use log-polar descriptor instead of HardNet | descriptor network · New dataset | FPR95: 0.98 -> 0.36 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- L2-Net — L2-Net: Deep Learning of Discriminative Patch Descriptor in Euclidean Space (CVPR 2017) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Explicit Spatial Encoding for Deep Local Descriptors (CVPR 2019 · similarity 0.89)
- Understanding and Improving Kernel Local Descriptors (IJCV 2018 · similarity 0.89)
- Local Descriptors Optimized for Average Precision (CVPR 2018 · similarity 0.88)
- Convolutional Patch Representations for Image Retrieval: An Unsupervised Approach (IJCV 2016 · similarity 0.88)
- Descriptor Distillation: A Teacher-Student-Regularized Framework for Learning Local Descriptors (IJCV 2024 · similarity 0.87)
- L2-Net: Deep Learning of Discriminative Patch Descriptor in Euclidean Space (CVPR 2017 · 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.
