Research map / Detection and segmentation / CNN features
Enhancing Deformable Local Features by Jointly Learning To Detect and Describe Keypoints
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 |
|---|---|---|---|
| Jointly learn keypoint detection and deformation-aware description | DALF · Deformable image matching, mean | MS: 0.45 -> 0.53 | Helped |
| Fuse distinctive backbone and invariant warper descriptors (ablation) | DALF · Deformable image matching, ablation | MS: 0.48 -> 0.53 | Helped |
| Add the non-rigid warper descriptor to the backbone features (ablation) | DALF · Deformable image matching, ablation | MS: 0.48 -> 0.53 | Helped |
| Use warper features only instead of backbone features only (ablation) | DALF · Deformable image matching, ablation | MS: 0.48 -> 0.51 | Helped |
| Use two-stage training rather than end-to-end training (ablation) | DALF with fused features · Deformable image matching, ablation | MMA: 0.79 -> 0.78 | Hurt |
| Use two-stage training rather than end-to-end training (ablation) | DALF with fused features · Deformable image matching, per-dataset MS | MS: Kinect1: 0.53 -> 0.55; Kinect2: 0.60 -> 0.63; DeSurT: 0.46 -> 0.50; Simulation: 0.50 -> 0.34 | Mixed |
| Add attention-based fusion to two-stage feature fusion (ablation) | DALF · Deformable image matching, ablation | MMA: 0.78 -> 0.80 | Helped |
| Add attention-based fusion to two-stage feature fusion (ablation) | DALF · Deformable image matching, ablation | MS: 0.53 -> 0.53 | No clear effect |
| Use jointly learned deformation-aware features | DALF · Deformable object retrieval | normalized area under curve: 98.57% -> 99.49% | Helped |
| Use DALF correspondences for non-rigid surface registration | ARAP registration · Non-rigid 3D surface registration @0.5cm | 3D accuracy: 42.8 -> 46.2 | Helped |
Closest papers by meaning
- Extracting Deformation-Aware Local Features by Learning to Deform (NeurIPS 2021 · similarity 0.91)
- Deformable Kernels: Adapting Effective Receptive Fields for Object Deformation (ICLR 2020 · similarity 0.87)
- Unsupervised Learning of Dense Shape Correspondence (CVPR 2019 · similarity 0.87)
- RDD: Robust Feature Detector and Descriptor using Deformable Transformer (CVPR 2025 · similarity 0.86)
- KP-RED: Exploiting Semantic Keypoints for Joint 3D Shape Retrieval and Deformation (CVPR 2024 · similarity 0.86)
- DV-Matcher: Deformation-based Non-rigid Point Cloud Matching Guided by Pre-trained Visual Features (CVPR 2025 · similarity 0.85)
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.