Research map / Label-efficient and robust learning / Recognition and long-tail
Structured Feature Similarity With Explicit Feature Map
Takumi Kobayashi
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 the proposed TENSOR similarity instead of L2 | SIFT image retrieval · Oxford building | mAP: 67.6 -> 73.2 | Improved |
| Apply the proposed feature map to CNN features | CNN features · Scene-15 | accuracy: 90.3±0.2 -> 91.2±0.5 | Improved |
| Apply the proposed feature map to CNN features | CNN features · SUN-397 | accuracy: 52.1±0.3 -> 54.3±0.3 | Improved |
| Use explicit feature mapping rather than pairwise similarity computation (ablation) | Proposed SIFT similarity · Keypoint matching, 4096 sample pairs | time per sample pair: 1146.5 -> 65.7 | Improved |
| Use TENSOR sub-features rather than VECTOR sub-features (ablation) | SSIM-based SIFT matching · Mikolajczyk and Schmid keypoint matching dataset | average precision | Improved |
| Preserve the intrinsic SIFT tensor layout rather than randomly permuting feature elements (ablation) | SSIM-based SIFT matching · Mikolajczyk and Schmid keypoint matching dataset | average precision | Improved |
| Use an additive SSIM formulation rather than the original multiplicative formulation (ablation) | TENSOR-structured SIFT matching · Mikolajczyk and Schmid keypoint matching dataset | average precision | Improved |
| Combine two of M, V, and C rather than using any single ingredient alone (ablation) | Sadd similarity · Mikolajczyk and Schmid keypoint matching dataset | average precision | Improved |
| Use the proposed similarity for image quality assessment rather than original SSIM | Image similarity assessment · TID2008 | Kendall rank correlation coefficient: 0.5768 -> 0.5489 | Hurt |
| Replace L2 with the proposed similarity for SURF-128 keypoint matching | SURF-128 · Mikolajczyk and Schmid keypoint matching dataset | average precision | Improved |
Closest papers by meaning
- Estimating the Number of Correct Matches Using Only Spatial Order (TPAMI 2018 · similarity 0.83)
- Progressive Large Scale-Invariant Image Matching in Scale Space (ICCV 2017 · similarity 0.82)
- Using Spatial Order to Boost the Elimination of Incorrect Feature Matches (CVPR 2016 · similarity 0.82)
- Local Similarity-Aware Deep Feature Embedding (NeurIPS 2016 · similarity 0.82)
- MESA: Matching Everything by Segmenting Anything (CVPR 2024 · similarity 0.80)
- Going Beyond Neural Network Feature Similarity: The Network Feature Complexity and Its Interpretation Using Category Theory (ICLR 2024 · similarity 0.80)
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.
