Pith. sign in

REVIEW 3 major objections 8 minor 33 references

Enhancing Visual Re-ranking through Denoising Nearest Neighbor Graph via Continuous CRF

T0 review · 3 major / 8 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Nearest-neighbor graph re-ranking improves when noisy edges are denoised by a continuous CRF that compares each node's similarity distribution over its local clique.

desk verdict A plausible, incremental graph-denoising preprocessor with a real but narrowly demonstrated SD term; the 'consistent gains' claim is overreaching and validation is tuned on the test set. read the letter →

arxiv 2412.13875 v2 pith:JVKXCE7R submitted 2024-12-18 cs.CV

Jaeyoon Kim , Yoonki Cho , Taeyoung Kim , Sung-Eui Yoon This is my paper · ORCID
classification cs.CV
keywords visualre-rankingnearestneighborgraphdenoisingcontinuousconditionalrandomfieldsstatisticaldistanceJeffreysdivergencediffusionprocessimageretrieval
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Visual re-ranking on nearest neighbor (NN) graphs is held back by noisy edges—links between images that are not actually similar. This paper claims that these edges can be cleaned before re-ranking by running a Continuous Conditional Random Field (C-CRF) on fully connected cliques built around each database image. The denoising uses a weight that combines Euclidean distance between CNN features with a statistical distance between softmax-normalized similarity distributions, so that hard negatives are identified by how the whole clique views them rather than by pairwise agreement. If the claim holds, the method is a plug-in pre-processing step that consistently improves three NN graph-based re-ranking approaches on ROxford and RParis without any fine-tuning; the offline diffusion baseline on ROxford Medium, for instance, rises from 69.9 to 76.1 mAP with ResNet features.

What carries the argument

The central object is a Continuous Conditional Random Field (C-CRF) defined separately on each fully connected clique $\mathcal{C}_p$ of $L$ nearest neighbors around a pivot image $I_p$. Its energy combines a unary potential keeping refined similarities close to initial ones and a pairwise potential whose weight is $w(f_i,f_j,\mathcal{S}_{\mathcal{C}_p}) = \exp\!\left(-\|f_i-f_j\|_2^2 / (2\sigma_d^2) - D_J(Q_i\|Q_j)^2 / (2\sigma_r^2)\right)$, where $Q_i$ is the softmax of $\ell^2$-normalized similarities from node $i$ to all other clique members and $D_J$ is Jeffreys divergence. The pairwise potential pulls refined values of nodes with high weight together, so the clique reaches a consensus that suppresses noisy edges. Because the energy is quadratic, the posterior is a multivariate Gaussian whose mean is the closed-form solution $\mu_p = \Sigma_p b_p$, computed once per clique with conjugate gradient; the resulting denoised similarities are averaged symmetrically and fed as the affinity to downstream re-ranking.

What would settle it

On a dataset where the initial similarity matrix is corrupted by randomly permuting the top-k neighbor lists (so that every node's similarity distribution is nearly uniform), the statistical distance would carry no information and the C-CRF-denoised affinity should yield the same retrieval mAP as the un-denoised reciprocal-NN baseline; observing a significant gap in such a controlled corruption experiment would test whether the claimed mechanism is actually driving the gains.

Watch

Extended reading notes

Core claim

The paper's central claim is that a C-CRF denoising pass over per-image cliques, with a weight function combining feature Euclidean distance and Jeffreys divergence of similarity-based distributions, consistently improves three different NN graph-based re-ranking methods (online diffusion, offline diffusion, and Explore-Exploit Graph Traversal) on standard landmark retrieval benchmarks. The key move is to replace pairwise reciprocity checks with a clique-wide consensus: each node is described by its softmax-normalized similarity distribution over the other clique members, and two nodes are considered conforming when that distribution is statistically close. The refined similarities come from the closed-form mean of a multivariate Gaussian C-CRF, computed offline, and are symmetrized into an affinity matrix that existing re-ranking methods consume unchanged. On the reported experiments, the largest gain is offline diffusion on ROxford Hard with ResNet, from 41.1 to 50.3 mAP.

Load-bearing premise

The statistical distance term assumes that a node's softmax-normalized similarity distribution over its clique is a reliable signature of which visual manifold it belongs to, even when the clique contains many noisy edges; if initial similarities are too noisy for those distributions to separate true neighbors from hard negatives, the C-CRF refinement has no useful signal and the denoised graph should not beat the original.

Editorial extensions

If this is right

  • Offline diffusion on ROxford Medium improves from 69.9 to 76.1 mAP with ResNet features, and the paper reports consistent gains for online diffusion and EGT as well.
  • Since the denoising is an offline pre-processing step, it can be added to existing retrieval pipelines without retraining or online overhead.
  • With noisy edges removed, diffusion-based re-ranking stays accurate at smaller k values, cutting online time complexity for a given quality target.
  • The improvement is stable over a wide range of clique sizes, which relieves some of the parameter sensitivity of diffusion methods.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The similarity-based distribution descriptor may generalize beyond NN graphs: any graph structure where nodes have a neighborhood distribution, such as k-reciprocal graphs or hierarchical navigable graphs, could benefit from the same clique-level Jeffreys-divergence weight.
  • Because SD compares whole distributions over the same clique rather than single feature distances, the approach may transfer to other consensus problems like outlier detection or clustering in high-dimensional feature spaces where hard negatives obscure pairwise distances.
  • A testable extension the paper does not explore is to make the bandwidths $\sigma_d$ and $\sigma_r$ per-clique adaptive; the current fixed values work on two datasets, but learned or data-dependent bandwidths could broaden applicability at the cost of the author's deliberate fine-tuning-free property.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 8 minor

Summary. This paper proposes a training-free preprocessing step for NN-graph-based visual re-ranking. For each database image, the method builds a fully connected clique of L nearest neighbors and refines the clique's similarity values by solving a continuous conditional random field with a unary potential that keeps the refined values close to the initial similarities and a pairwise potential that pulls together nodes deemed similar. The pairwise weight combines a Gaussian kernel on feature Euclidean distance and Jeffreys divergence between softmax-normalized similarity distributions (SBD), intended to downweight noisy edges. The refined affinities are symmetrized and then fed to existing re-ranking methods, namely online diffusion, offline diffusion, and EGT. On ROxford and RParis, the paper reports mAP gains for most settings.

Significance. If the results are reproducible, the method is a useful plug-in that can improve standard diffusion and graph-traversal re-rankers without fine-tuning, and the closed-form C-CRF inference is mathematically clean. The evaluation uses standard benchmarks, and the method does not fit ground truth. The most impressive reported gain is offline diffusion on ROxford Medium with ResNet, from 69.9 to 76.1 mAP in Table 3. The significance is moderate: the contribution is a preprocessing/affinity-denoising module rather than a new re-ranking model, and the distinctive statistical-distance mechanism needs more direct evidence than the current mAP-level ablations provide.

major comments (3)
  1. [Abstract; Sec. 4.2; Tables 1, 2] The abstract and Sec. 4.2 claim that the method 'consistently improves three different NN graph-based re-ranking approaches,' but Tables 1 and 2 contain two degradations: EGT with VGG drops 0.8 mAP on ROxford Easy (88.3 to 87.5) and EGT with ResNet drops 0.1 mAP on RParis Easy (92.8 to 92.7). Please replace 'consistently' with a qualified statement such as 'generally' or 'in most settings,' and briefly discuss these failure cases or show that they are within run-to-run variation.
  2. [Sec. 3.1, Eqs. (9)-(11); Table 3] The distinctive claim of the paper is that the statistical-distance (SD) term is robust 'even in this extreme case' where a clique contains many noisy edges (Sec. 3.1). This mechanism is not directly tested. Table 3 reports only final mAP; for ResNet, ED alone gives 75.5 vs 76.1 for ED+SD, so the additional signal from SD is small, and although SD alone is stronger on VGG (72.7 vs 70.9 for ED alone), the mAP-level ablation cannot localize the benefit to specific noisy edges. A hard negative is by construction visually similar to the query and to many true positives, so its similarity-based PMF over the clique can be dominated by the same true-positive cluster as a true neighbor's PMF, making D_J small and up-weighting the edge. Please add an edge-level or synthetic experiment that directly measures whether SD separates true neighbors from hard negatives in noisy cliques (e.g., precision/recall of retained edges against ground truth, or weight distributions for true and false edges). This would also test the load-bearing premise of the method.
  3. [Sec. 4.1; Fig. 4] Several hyperparameters are selected on the test benchmarks themselves. Sec. 4.1 states that clique sizes of 1,000 and 500 are 'empirically chosen' for ROxford and RParis, and Fig. 4 justifies the choice by plotting mAP versus clique size on the Medium protocol of the same datasets. The paper does not state whether the parameters sigma_d, sigma_r, and L were chosen on a validation split or on the test sets, and Fig. 4 indicates that L was chosen on the test set. This makes the reported gains optimistic and weakens the comparison to baselines whose parameters are fixed by their original publications. Please select parameters on a validation split, or report sensitivity without selecting the best configuration on test, and then report test results for the chosen configuration.
minor comments (8)
  1. [Table 1, VGG section] The NN-Search value in the Hard column appears truncated as '32.'; please correct the value.
  2. [Sec. 2, after Eq. (4)] The phrase 'the values inf * contain' should read 'the values in f* contain'.
  3. [Eq. (9)] Define \hat{s}_{i,k} explicitly as the l2-normalization of s_{i,k} over the clique dimension; the current wording is ambiguous.
  4. [Tables 1, 2] The label 'Offline adiffusion' appears to be a typo for 'Offline diffusion'.
  5. [Sec. 2, Eqs. (3)-(5)] The initial state is denoted v0 in Eq. (3) but f0 in Eq. (4) and the following text; please use a single notation.
  6. [Sec. 4.2] The 'average improvements' of 4.6 mAP for ROxford and 1.7 mAP for RParis are not defined; please specify the aggregation over methods and protocols.
  7. [Reference [23]] The reference for Jeffreys divergence cites 'Scientific inference' (1973); the standard source is Jeffreys (1946) 'An Invariant Form for the Prior Probability in Estimation Problems' or his 'Theory of Probability'. Please update the citation.
  8. [Table 3] The baseline row in Table 3 is not labeled; add a label such as 'Offline diffusion' for clarity.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the C-CRF denoising is an unsupervised preprocessing transformation evaluated on external benchmarks, with no fitted parameter renamed as a prediction.

full rationale

The paper proposes a C-CRF-based refinement of the initial affinity matrix. The output y* is the closed-form solution of a quadratic energy (Eqs. 6-14) whose unary term pulls y toward the initial similarities and whose pairwise term couples nodes via a weight w that combines CNN feature Euclidean distance and Jeffreys divergence of softmax-normalized similarity distributions (Eqs. 9-11). This is a well-defined smoothing/filtering operation: the input is the initial similarity matrix, and the output is a regularized version of the same matrix. No ground-truth labels are used, no parameter is fitted to the retrieval targets, and the claim is not a derivation but an empirical improvement in downstream mAP on ROxford/RParis. The SBD is indeed computed from the same initial similarities being refined, but that is a design assumption about what signal separates true neighbors from hard negatives, not a circular derivation: the paper never defines the denoising target in terms of the refined output, nor does it predict a quantity that was used as a fitting target. The only self-citation is [7] in the related-work survey of geometric verification, and it is not load-bearing. The blue entries in Tables 1-2 contradict the word 'consistently' in a few configurations, but that is an empirical inconsistency, not circularity. No uniqueness theorem is invoked, and no result is imported from the authors' prior work to force the choice. Parameters such as clique size and sigma are chosen empirically from the datasets, which may raise overfitting concerns, but that is not circularity. The derivation chain is therefore self-contained; the appropriate circularity score is 0.

Assumptions & free parameters 6 free parameters · 4 assumptions · 1 invented entities

The central mechanism depends on hand-chosen hyperparameters (sigma_d, sigma_r, L, alpha, beta) and on the SBD assumption that similarity distributions remain informative despite noisy edges. No new physical entities are introduced; SBD is a mathematical descriptor.

free parameters (6)
  • sigma_d (feature distance bandwidth) = 0.8 (VGG), 0.9 (ResNet)
    Controls the Gaussian kernel on Euclidean distance between CNN features in Eq. 11; chosen empirically with no validation split stated.
  • sigma_r (statistical distance bandwidth) = 2e-4 (VGG), 3.5e-4 (ResNet)
    Controls the Gaussian kernel on Jeffreys divergence in Eq. 11; chosen empirically.
  • Clique size L = 1,000 (ROxford), 500 (RParis)
    Defines the set of nearest neighbors forming the clique in Sec. 3.1; Figure 4 is used to pick sizes, so selection depends on test set performance.
  • alpha (unary potential weight) = 1
    Balance between unary and pairwise potentials in Eq. 7; fixed by hand, not derived.
  • beta (pairwise potential weight) = 0.1
    Balance between unary and pairwise potentials in Eq. 7; fixed for all experiments.
  • k for affinity construction = varied 10 to 70 in Fig. 3; otherwise follows baselines
    Sparsity of the NN graph influences diffusion complexity; the paper shows its method reduces sensitivity to k, but the value is still a free parameter.
assumptions (4)
  • standard math The C-CRF multivariate Gaussian form and inference y* = mean = precision^{-1} b (Eqs. 12-14) from Radosavljevic et al. are correct and applicable to clique subgraphs.
    Sec. 3.1 relies on this prior result without reproving it.
  • ad hoc to paper Softmax-normalized similarity distributions Q_i (SBD) are reliable statistical descriptors of node identity even when the clique contains many noisy edges.
    Introduced in Sec. 3.1 as the basis of the statistical distance; no independent evidence except benchmark performance.
  • domain assumption Fully connected cliques of size L around every pivot contain enough collective information to detect noisy edges, and processing cliques independently is sufficient.
    Sec. 3.1 states 'we choose a clique... where we can have ample information such that we can identify noisy edges'; no theorem guarantees this.
  • domain assumption Initial affinities via reciprocity check (Eq. 2) provide a reasonable starting point for C-CRF refinement.
    Sec. 2 constructs A from min of truncated similarities; the unary term treats these as noisy observations to be preserved.
invented entities (1)
  • Similarity-Based Distribution (SBD)
    purpose: Acts as a per-node statistical descriptor in the pairwise weight function, enabling hard-negative detection via Jeffreys divergence.
    Defined in Eq. 9 as l2-normalized similarities followed by softmax; it is a mathematical construct internal to the method, with no separate falsifiable handle.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Enhancing Visual Re-ranking through Denoising Nearest Neighbor Graph via Continuous CRF." pith.science (2026). https://pith.science/paper/JVKXCE7R

@misc{pith2026241213875,
  author       = {Pith},
  title        = {Pith review of: Enhancing Visual Re-ranking through Denoising Nearest Neighbor Graph via Continuous CRF},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JVKXCE7R}},
  note         = {Machine review of arXiv:2412.13875}
}
read the original abstract

Nearest neighbor (NN) graph based visual re-ranking has emerged as a powerful approach for improving retrieval accuracy, offering the advantages of effectively exploring high-dimensional manifolds without requiring additional fine-tuning. However, the effectiveness of NN graph-based re-ranking is fundamentally constrained by the quality of its edge connectivity, as incorrect connections between dissimilar (negative) images frequently occur. This is known as a noisy edge problem, which hinders the re-ranking performance of existing techniques and limits their potential. To remedy this issue, we propose a complementary denoising method based on Continuous Conditional Random Fields (C-CRF) that leverages statistical distances derived from similarity-based distributions. As a pre-processing step for enhancing NN graph-based retrieval, our approach constructs fully connected cliques around each anchor image and employs a novel statistical distance metric to robustly alleviate noisy edges before re-ranking while achieving efficient processing through offline computation. Extensive experimental results demonstrate that our method consistently improves three different NN graph-based re-ranking approaches, yielding significant gains in retrieval accuracy.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

33 extracted references · 32 canonical work pages

  1. [1]

    Enhancing Visual Re-ranking through Denoising Nearest Neighbor Graph via Continuous CRF

    INTRODUCTION Visual retrieval systems typically perform nearest neighbor search with visual descriptors; however, they face fundamen- tal challenges when handling non-linear manifold structures in high-dimensional feature spaces [1, 2]. To address this issue, re-ranking process has emerged as a crucial compo- nent in visual retrieval systems, refining ini...

    work page Pith review arXiv 2025

  2. [2]

    The diffusion process in visual re-ranking constructs revised sim- ilarity considering manifold from initial similarity computed from all images in the database

    PRELIMINARIES In this section, we summarize the diffusion process to show how the NN-graph can be applied for visual re-ranking. The diffusion process in visual re-ranking constructs revised sim- ilarity considering manifold from initial similarity computed from all images in the database. Generally, this process con- structs an undirected graph structure...

    work page

  3. [3]

    2) with k- NN of each database image to ameliorate noisy edges whose vertices are not positive, i.e., similar images

    APPROACH Diffusion methods [14, 22] for image retrieval adopted a sim- ple approach of using the reciprocity check (Eq. 2) with k- NN of each database image to ameliorate noisy edges whose vertices are not positive, i.e., similar images. We, however, (a) (b) (c) Fig. 2: An example of clique-based denoising;k = 7 and L =

    work page

  4. [4]

    w/ ED” and “w/ SD

    EXPERIMENTS In this section, we explore the application of our denoising approach to landmark retrieval with three visual re-ranking methods. Table 1: mean Average Precision (mAP) comparisons against other re-ranking methods using the NN-graph for ROxford. ResNet and VGG for extracting global features are fine-tuned networks. Improvements without online-t...

    work page 2000

  5. [5]

    Our method leverages fully con- nected cliques and a novel statistical distance metric to ro- bustly alleviate noisy edges

    CONCLUSION We propose a C-CRF-based denoising approach to enhance NN graph-based re-ranking. Our method leverages fully con- nected cliques and a novel statistical distance metric to ro- bustly alleviate noisy edges. Experimental results demon- strate consistent improvements over existing re-ranking ap- proaches, confirming the effectiveness of our method...

    work page 2023

  6. [6]

    Local features and visual words emerge in activations,

    Oriane Sim ´eoni, Yannis Avrithis, and Ondrej Chum, “Local features and visual words emerge in activations,” in CVPR, 2019

    work page 2019

  7. [7]

    Deep image retrieval: Learning global repre- sentations for image search,

    Albert Gordo, Jon Almaz ´an, J´erome Revaud, and Diane Larlus, “Deep image retrieval: Learning global repre- sentations for image search,” in ECCV, 2016

    work page 2016

  8. [8]

    Recent ad- vance in content-based image retrieval: A literature sur- vey,

    Wengang Zhou, Houqiang Li, and Qi Tian, “Recent ad- vance in content-based image retrieval: A literature sur- vey,” arXiv preprint arXiv:1706.06064, 2017

    arXiv 2017

Show all 33 references
  1. [9]

    Regularized diffusion process on bidirectional context for object retrieval,

    Song Bai, Xiang Bai, Qi Tian, and Longin Jan Latecki, “Regularized diffusion process on bidirectional context for object retrieval,” TPAMI, 2019

    2019

  2. [10]

    (b) shows a clique of the pivot, and all nodes in the clique are fully connected

    (a) shows initial edges of a yellow pivot node, and red edges are noisy edges. (b) shows a clique of the pivot, and all nodes in the clique are fully connected. (c) represents edges denoised by C-CRF, which is constructed from the clique. found that the reciprocity check is li...

  3. [11]

    Object retrieval with large vo- cabularies and fast spatial matching,

    James Philbin, Ondrej Chum, Michael Isard, Josef Sivic, and Andrew Zisserman, “Object retrieval with large vo- cabularies and fast spatial matching,” in CVPR, 2007

    2007

  4. [12]

    Three things everyone should know to improve object re- trieval,

    Relja Arandjelovi ´c and Andrew Zisserman, “Three things everyone should know to improve object re- trieval,” in CVPR, 2012

    2012

  5. [13]

    Topological ransac for in- stance verification and retrieval without fine-tuning,

    Guoyuan An, Ju-hyeong Seon, Inkyu An, Yuchi Huo, and Sung-Eui Yoon, “Topological ransac for in- stance verification and retrieval without fine-tuning,” in NeurIPS, 2023

    2023

  6. [14]

    Total recall: Automatic query expansion with a generative feature model for ob- ject retrieval,

    Ondrej Chum, James Philbin, Josef Sivic, Michael Is- ard, and Andrew Zisserman, “Total recall: Automatic query expansion with a generative feature model for ob- ject retrieval,” in ICCV, 2007

    2007

  7. [15]

    Fine-tuning CNN image retrieval with no human an- notation,

    Filip Radenovic, Giorgos Tolias, and Ondrej Chum, “Fine-tuning CNN image retrieval with no human an- notation,” TPAMI, 2019

    2019

  8. [16]

    Attention-based query expansion learning,

    Albert Gordo, Filip Radenovic, and Tamara Berg, “Attention-based query expansion learning,” in ECCV, 2020

    2020

  9. [17]

    Contextual similarity aggregation with self-attention for visual re-ranking,

    Jianbo Ouyang, Hui Wu, Min Wang, Wengang Zhou, and Houqiang Li, “Contextual similarity aggregation with self-attention for visual re-ranking,” in NeurIPS, 2021

    2021

  10. [18]

    Ranking on data manifolds,

    Dengyong Zhou, Jason Weston, Arthur Gretton, Olivier Bousquet, and Bernhard Sch ¨olkopf, “Ranking on data manifolds,” in NIPS, 2003

    2003

  11. [19]

    Diffusion pro- cesses for retrieval revisited,

    Michael Donoser and Horst Bischof, “Diffusion pro- cesses for retrieval revisited,” in CVPR, 2013

    2013

  12. [20]

    Efficient diffusion on re- gion manifolds: Recovering small objects with compact CNN representations,

    Ahmet Iscen, Giorgos Tolias, Yannis Avrithis, Teddy Furon, and Ondrej Chum, “Efficient diffusion on re- gion manifolds: Recovering small objects with compact CNN representations,” in CVPR, 2017

    2017

  13. [21]

    Explore-exploit graph traversal for image re- trieval,

    Cheng Chang, Guangwei Yu, Chundi Liu, and Maksims V olkovs, “Explore-exploit graph traversal for image re- trieval,” in CVPR, 2019

    2019

  14. [22]

    Object retrieval with image graph traversal-based re-ranking,

    Siyuan Qi and Yupin Luo, “Object retrieval with image graph traversal-based re-ranking,” Signal Processing: Image Communication, 2016

    2016

  15. [23]

    Locally constrained diffusion process on locally densified distance spaces with applications to shape retrieval,

    Xingwei Yang, Suzan K ¨oknar-Tezel, and Longin Jan Latecki, “Locally constrained diffusion process on locally densified distance spaces with applications to shape retrieval,” in CVPR, 2009

    2009

  16. [24]

    An introduction to conditional random fields for relational learning,

    C Sutton, “An introduction to conditional random fields for relational learning,” Introduction to statistical rela- tional learning, 2007

    2007

  17. [25]

    Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs,

    Liang-Chieh Chen, George Papandreou, Iasonas Kokki- nos, Kevin Murphy, and Alan L Yuille, “Deeplab: Semantic image segmentation with deep convolutional nets, atrous convolution, and fully connected crfs,” TPAMI, 2017

    2017

  18. [26]

    Global ranking using continuous conditional random fields,

    Tao Qin, Tie-Yan Liu, Xu-Dong Zhang, De-Sheng Wang, and Hang Li, “Global ranking using continuous conditional random fields,” in NIPS, 2008

    2008

  19. [27]

    Continuous conditional random fields for efficient regression in large fully con- nected graphs,

    Kosta Ristovski, Vladan Radosavljevic, Slobodan Vucetic, and Zoran Obradovic, “Continuous conditional random fields for efficient regression in large fully con- nected graphs,” in AAAI, 2013

    2013

  20. [28]

    Efficient image retrieval via de- coupling diffusion into online and offline processing,

    Fan Yang, Ryota Hinami, Yusuke Matsui, Steven Ly, and Shin’ichi Satoh, “Efficient image retrieval via de- coupling diffusion into online and offline processing,” in AAAI, 2019

    2019

  21. [29]

    Harold Jeffreys, Scientific inference, Cambridge Uni- versity Press, 1973

    1973

  22. [30]

    Continuous conditional random fields for regression in remote sensing,

    Vladan Radosavljevic, Slobodan Vucetic, and Zoran Obradovic, “Continuous conditional random fields for regression in remote sensing,” in ECAI, 2010

    2010

  23. [31]

    An introduction to the conjugate gradient method without the agonizing pain,

    Jonathan Richard Shewchuk et al., “An introduction to the conjugate gradient method without the agonizing pain,” 1994

    1994

  24. [32]

    Revisiting oxford and paris: Large-scale image retrieval benchmarking,

    Filip Radenovic, Ahmet Iscen, Giorgos Tolias, Yannis Avrithis, and Ondrej Chum, “Revisiting oxford and paris: Large-scale image retrieval benchmarking,” in CVPR, 2018

    2018

  25. [33]

    Genetic algorithms for the optimiza- tion of diffusion parameters in content-based image re- trieval,

    Federico Magliani, Laura Sani, Stefano Cagnoni, and Andrea Prati, “Genetic algorithms for the optimiza- tion of diffusion parameters in content-based image re- trieval,” arXiv preprint arXiv:1908.06896, 2019

    1908 arXiv

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.