Student or educator? Sign up with your school email and get 6 months of Pro free.
Cited by 1,472

The DOI matches its Crossref record; 13 of the 13 references it cites were found in a registry or an index.

Fine-Tuning CNN Image Retrieval with No Human Annotation

Open access
Radenovic, Filip, Tolias, Giorgos, Chum, Ondrej·2019·IEEE Transactions on Pattern Analysis and Machine Intelligence

Abstract

Image descriptors based on activations of Convolutional Neural Networks (CNNs) have become dominant in image retrieval due to their discriminative power, compactness of representation, and search efficiency. Training of CNNs, either from scratch or fine-tuning, requires a large amount of annotated data, where a high quality of annotation is often crucial. In this work, we propose to fine-tune CNNs for image retrieval on a large collection of unordered images in a fully automated manner. Reconstructed 3D models obtained by the state-of-the-art retrieval and structure-from-motion methods guide the selection of the training data. We show that both hard-positive and hard-negative examples, selected by exploiting the geometry and the camera positions available from the 3D models, enhance the performance of particular-object retrieval. CNN descriptor whitening discriminatively learned from the same training data outperforms commonly used PCA whitening. We propose a novel trainable Generalized-Mean (GeM) pooling layer that generalizes max and average pooling and show that it boosts retrieval performance. Applying the proposed method to the VGG network achieves state-of-the-art performance on the standard benchmarks: Oxford Buildings, Paris, and Holidays datasets.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/tpami.2018.2846566

Cites

13 of 13 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 13 references

Research works printed but not linked: 51

  1. srivastava, 2014, Dropout: A simple way to prevent neural networks from overfitting, J Mach Learn Res, 15, 1929
  2. 10.1109/CVPR.2005.202 · DOI · Paper page
  3. 10.1016/j.patcog.2014.04.007 · DOI
  4. 10.1109/CVPR.2007.383172 · DOI · Paper page
  5. 10.1007/978-3-319-10599-4_25 · DOI
  6. 10.1109/CVPR.2013.207 · DOI · Paper page
  7. 10.1145/2671188.2749366 · DOI
  8. danfeng, 2011, Hello neighbor: Accurate object retrieval with k-reciprocal nearest neighbors, Proc IEEE Conf Comput Vis Pattern Recognit, 777
  9. 10.1109/CVPR.2011.5995601 · DOI · Paper page
  10. 10.1109/TPAMI.2013.237 · DOI
  11. 10.1007/978-3-642-33712-3_2 · DOI
  12. 10.1007/978-3-642-15561-1_27 · DOI · Paper page
  13. 10.1145/2001269.2001293 · DOI · Paper page
  14. 10.1007/978-3-642-41062-8_2 · DOI
  15. 10.1109/CVPR.2010.5540009 · DOI · Paper page
  16. mikulík, 2014, Efficient image detail mining, Proc Asian Conf Comput Vis, 118
  17. 10.1007/s11263-017-1016-8 · DOI · Paper page
  18. 10.1007/978-3-642-15552-9_57 · DOI · Paper page
  19. 10.1145/2733373.2807412 · DOI · Paper page
  20. 10.1109/TPAMI.2011.235 · DOI · Paper page
  21. 10.1109/CVPR.2008.4587635 · DOI · Paper page
  22. 10.1007/s11263-010-0363-5 · DOI
  23. jégou, 2012, Negative evidences and co-occurences in image retrieval: The benefit of PCA and whitening, Proc Eur Conf Comput Vis, 774
  24. 10.1007/978-3-319-46466-4_15 · DOI · Paper page
  25. 10.1007/978-3-319-46448-0_1 · DOI · Paper page
  26. lee, 2016, Generalizing pooling functions in convolutional neural networks: Mixed, gated, and tree, Proc Int Conf Artif Intell Statist, 464
  27. 10.1109/CVPR.2016.517 · DOI
  28. 10.1007/s11263-012-0600-1 · DOI
  29. 10.1109/CVPR.2012.6248018 · DOI · Paper page
  30. radenovi?, 2016, From dusk till dawn: Modeling in the dark, Proc IEEE Conf Comput Vis Pattern Recognit, 5488
  31. 10.5244/C.23.91 · DOI · Paper page
  32. 10.1109/CVPR.2015.7298636 · DOI · Paper page
  33. 10.1109/ICCV.2007.4408871 · DOI
  34. morère, 2017, Nested invariance pooling and RBM hashing for image instance retrieval, Proc ACM Int Conf Multimedia Retrieval, 260, 10.1145/3078971.3078987 · DOI
  35. babenko, 2015, Aggregating deep convolutional features for image retrieval, Proc Int Conf Comput Vis, 1269
  36. 10.1007/978-3-319-46604-0_48 · DOI · Paper page
  37. 10.1109/CVPR.2006.100 · DOI · Paper page
  38. 10.1007/978-3-319-10590-1_54 · DOI · Paper page
  39. 10.1007/978-3-319-10590-1_38 · DOI · Paper page
  40. arandjelovic, 2016, NetVLAD: CNN architecture for weakly supervised place recognition, Proc IEEE Conf Comput Vis Pattern Recognit, 5297
  41. 10.1109/CVPR.2015.7299148 · DOI
  42. 10.1109/TPAMI.2009.166 · DOI
  43. 10.1109/ICCV.2013.432 · doi:10.1109/iccv.2013.432
  44. 10.1109/CVPRW.2015.7301270 · DOI · Paper page
  45. 10.1109/CVPRW.2014.131 · DOI · Paper page
  46. 10.1007/978-3-319-10584-0_26 · DOI · Paper page
  47. 10.1145/2911996.2912061 · DOI
  48. 10.1109/ICCV.2017.374 · DOI · Paper page
  49. 10.1109/CVPR.2014.180 · DOI · Paper page
  50. 10.1109/CVPR.2014.242 · DOI · Paper page
  51. 10.1007/978-3-319-24261-3_7 · DOI · Paper page

Cited by

See all 1,472

Metadata

Authors
Radenovic, F., Tolias, G., & Chum, O.
Venue
IEEE Transactions on Pattern Analysis and Machine Intelligence
Published
2019-07-01
Volume
41
Issue
7
Pages
1655-1668
ISSN
  • 0162-8828 (ISSN-L)
  • 2160-9292
  • 1939-3539
OpenAlex
W2767621619
Cited by
1,472
References
13
FWCI
53.46
Type
journal article
Language
en
Open access
green
Funders
MSMT (LL1303 ERC-CZ)

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Fine-Tuning CNN Image Retrieval with No Human Annotation. https://app.cochat.ai/papers/10.1109/tpami.2018.2846566
The paper
Radenovic, F., Tolias, G., & Chum, O. (2019). Fine-Tuning CNN Image Retrieval with No Human Annotation. IEEE Transactions on Pattern Analysis and Machine Intelligence. https://doi.org/10.1109/tpami.2018.2846566

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.