Abstract
State of the art image retrieval performance is achieved with CNN features and manifold ranking using a k-NN similarity graph that is pre-computed off-line. The two most successful existing approaches are temporal filtering, where manifold ranking amounts to solving a sparse linear system online, and spectral filtering, where eigen-decomposition of the adjacency matrix is performed off-line and then manifold ranking amounts to dot-product search online. The former suffers from expensive queries and the latter from significant space overhead. Here we introduce a novel, theoretically well-founded hybrid filtering approach allowing full control of the space-time trade-off between these two extremes. Experimentally, we verify that our hybrid method delivers results on par with the state of the art, with lower memory demands compared to spectral filtering approaches and faster compared to temporal filtering.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
- 2.
“Temporal” stems from conventional signal processing where signals are functions of “time”; while “spectral” is standardized also in graph signal processing.
- 3.
References
Arandjelovic, R., Zisserman, A.: Three things everyone should know to improve object retrieval. In: CVPR, June 2012
Bai, S., Bai, X., Tian, Q., Latecki, L.J.: Regularized diffusion process on bidirectional context for object retrieval. IEEE Trans. PAMI (2018)
Bai, S., Zhou, Z., Wang, J., Bai, X., Jan Latecki, L., Tian, Q.: Ensemble diffusion for retrieval. In: ICCV (2017)
Chum, O., Mikulik, A., Perdoch, M., Matas, J.: Total recall II: query expansion revisited. In: CVPR, June 2011
Chum, O., Philbin, J., Sivic, J., Isard, M., Zisserman, A.: Total recall: automatic query expansion with a generative feature model for object retrieval. In: ICCV, October 2007
Chung, F.R.: Spectral Graph Theory, vol. 92. American Mathematical Society (1997)
Delvinioti, A., Jégou, H., Amsaleg, L., Houle, M.: Image retrieval with reciprocal and shared nearest neighbors. In: VISAPP, January 2014
Donoser, M., Bischof, H.: Diffusion processes for retrieval revisited. In: CVPR (2013)
Gordo, A., Almazan, J., Revaud, J., Larlus, D.: End-to-end learning of deep visual representations for image retrieval. IJCV 124(2) (2017)
Hackbusch, W.: Iterative Solution of Large Sparse Systems of Equations. Springer, New York (1994). https://doi.org/10.1007/978-1-4612-4288-8
Iscen, A., Avrithis, Y., Tolias, G., Furon, T., Chum, O.: Fast spectral ranking for similarity search. In: CVPR (2018)
Iscen, A., Tolias, G., Avrithis, Y., Furon, T., Chum, O.: Efficient diffusion on region manifolds: recovering small objects with compact CNN representations. In: CVPR (2017)
Jégou, H., Harzallah, H., Schmid, C.: A contextual dissimilarity measure for accurate and efficient image search. In: CVPR, June 2007
Joly, A., Buisson, O.: Logo retrieval with a contrario visual query expansion. In: ACM Multimedia, October 2009
Kalantidis, Y., Mellina, C., Osindero, S.: Cross-dimensional weighting for aggregated deep convolutional features. arXiv preprint arXiv:1512.04065 (2015)
Philbin, J., Chum, O., Isard, M., Sivic, J., Zisserman, A.: Object retrieval with large vocabularies and fast spatial matching. In: CVPR, June 2007
Philbin, J., Chum, O., Isard, M., Sivic, J., Zisserman, A.: Lost in quantization: improving particular object retrieval in large scale image databases. In: CVPR, June 2008
Qin, D., Gammeter, S., Bossard, L., Quack, T., Van Gool, L.: Hello neighbor: accurate object retrieval with k-reciprocal nearest neighbors. In: CVPR (2011)
Radenović, F., Iscen, A., Tolias, G., Avrithis, Y., Chum, O.: Revisiting Oxford and Paris: large-scale image retrieval benchmarking. In: CVPR (2018)
Radenović, F., Tolias, G., Chum, O.: Fine-tuning CNN image retrieval with no human annotation. IEEE Trans. PAMI (2018)
Shen, X., Lin, Z., Brandt, J., Wu, Y.: Spatially-constrained similarity measure for large-scale object retrieval. IEEE Trans. PAMI 36(6), 1229–1241 (2014)
Sivic, J., Zisserman, A.: Video Google: a text retrieval approach to object matching in videos. In: ICCV (2003)
Tolias, G., Jégou, H.: Visual query expansion with or without geometry: refining local descriptors by feature aggregation. Pattern Recogn. 47(10), 3466–3476 (2014)
Tolias, G., Sicre, R., Jégou, H.: Particular object retrieval with integral max-pooling of CNN activations. ICLR (2016)
Tong, H., Faloutsos, C., Pan, J.Y.: Fast random walk with restart and its applications. In: Proceedings of the IEEE International Conference on Data Mining, pp. 613–622 (2006)
Trefethen, L.N., Bau III, D.: Numerical Linear Algebra. SIAM (1997)
Zhang, S., Yang, M., Cour, T., Yu, K., Metaxas, D.N.: Query specific fusion for image retrieval. In: Fitzgibbon, A., Lazebnik, S., Perona, P., Sato, Y., Schmid, C. (eds.) ECCV 2012. LNCS, pp. 660–673. Springer, Heidelberg (2012). https://doi.org/10.1007/978-3-642-33709-3_47
Zhou, D., Weston, J., Gretton, A., Bousquet, O., Schölkopf, B.: Ranking on data manifolds. In: NIPS (2003)
Acknowledgements
This work was supported by MSMT LL1303 ERC-CZ grant and the OP VVV funded project CZ.02.1.01/0.0/0.0/16_019/0000765 “Research Center for Informatics”.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
A General Derivation
A General Derivation
The derivation of our algorithm in Sect. 4.1 applies only to the particular function (filter) \(h_\alpha \) (3). Here, as in [11], we generalize to a much larger class of functions, that is, any function h that has a series expansion
We begin with the same eigenvalue decomposition (6) of \(\mathcal {W}\) and, assuming that \(h(\mathcal {W})\) converges absolutely, its corresponding decomposition
similar to (8), where \(U_1\), \(U_2\) have the same orthogonality properties (5).
Again, the first term is exactly the low-rank approximation that is used by spectral filtering, and the second is the approximation error
Again, we have used the series expansion (24) of h in (27) and (29). Now, Eq. (28) is due to the fact that
for \(i \ge 1\), as can be verified by induction, while for \(i = 0\),
In both (30) and (31) we have used the orthogonality properties (5).
Finally, combining (25), (29) and (6), we have proved the following.
Theorem 2
Assuming the series expansion (24) of transfer function h and the eigenvalue decomposition (6) of the symmetrically normalized adjacency matrix \(\mathcal {W}\), and given that \(h(\mathcal {W})\) converges absolutely, it is decomposed as
where
for \(n \times n\) real symmetric matrix A. For \(h = h_\alpha \) and for \(x \in [-1,1]\) in particular, \(g_\alpha (x) \mathrel {{\text {:=}}}h_\alpha (x) - h_\alpha (0) = (1 - \alpha ) \alpha x / (1 - \alpha x)\).
This general derivation explains where the general definition of function g (33) is coming from in (16) corresponding to our treatment of \(h_\alpha \) in Sect. 4.1.
Rights and permissions
Copyright information
© 2019 Springer Nature Switzerland AG
About this paper
Cite this paper
Iscen, A., Avrithis, Y., Tolias, G., Furon, T., Chum, O. (2019). Hybrid Diffusion: Spectral-Temporal Graph Filtering for Manifold Ranking. In: Jawahar, C., Li, H., Mori, G., Schindler, K. (eds) Computer Vision – ACCV 2018. ACCV 2018. Lecture Notes in Computer Science(), vol 11362. Springer, Cham. https://doi.org/10.1007/978-3-030-20890-5_20
Download citation
DOI: https://doi.org/10.1007/978-3-030-20890-5_20
Published:
Publisher Name: Springer, Cham
Print ISBN: 978-3-030-20889-9
Online ISBN: 978-3-030-20890-5
eBook Packages: Computer ScienceComputer Science (R0)