Pith. sign in

REVIEW 2 major objections 5 minor 37 references

ViSiL: Fine-grained Spatio-Temporal Video Similarity Learning

T0 review · 2 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read ViSiL claims video similarity should be computed from refined frame-to-frame similarity matrices built on regional CNN features, not from global video embeddings, and reports large gains across four retrieval tasks.

desk verdict ViSiL is a genuinely new and mostly well-evaluated video similarity architecture, with one benchmark comparison (EVVE) that doesn't support the headline claim. read the letter →

arxiv 1908.07410 v1 pith:GH4QI74V submitted 2019-08-20 cs.CV cs.IR

Giorgos Kordopatis-Zilos , Symeon Papadopoulos , Ioannis Patras , Ioannis Kompatsiaris This is my paper · ORCID
classification cs.CVcs.IR
keywords videoretrievalsimilaritylearningnear-duplicatefine-grainedincidentframe-to-framespatio-temporaltripletlossChamfer
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

The paper sets out to show that video-to-video similarity is better computed from fine-grained frame-to-frame similarity than from a single embedding of the whole video. It introduces ViSiL, a network that first scores every pair of frames using region-level CNN features, then passes the resulting similarity matrix through a small CNN and reduces it to one video-level score. This design preserves both spatial detail within frames and temporal patterns across frames, which global video descriptors throw away. On four retrieval problems—near-duplicate, incident/event, and action retrieval—the paper reports consistent gains over prior methods, often by a large margin.

What carries the argument

The load-bearing object is the frame-to-frame similarity matrix between all pairs of frames of two videos. It is built by decomposing each frame into $N \times N$ region vectors from intermediate CNN layers, weighting regions by a learned attention vector, computing the Tensor Dot product between region vectors of every frame pair, and applying Chamfer Similarity (the average, over one frame's regions, of each region's maximum similarity to the other frame's regions). A four-layer CNN then processes the matrix and a final Chamfer Similarity yields the video-to-video score, trained with triplet loss plus a regularization that keeps pre-activation values inside the hard-tanh range. The matrix is the carrier of the spatio-temporal information: regional matching supplies spatial invariance, while the CNN learns temporal patterns such as diagonals and segment-level blocks.

What would settle it

If an independent re-annotation of the top-ranked supposedly false positives in FIVR-200K showed they are genuinely irrelevant, or if rerunning the comparison on the original uncorrected CC_WEB_VIDEO labels let an existing method such as the circulant temporal encoding beat ViSiL, the claimed superiority would be in doubt.

Watch

Extended reading notes

Core claim

ViSiL's central claim is that comparison before aggregation is what matters: regional CNN features from intermediate layers are whitened, attention-weighted, and matched with Tensor Dot plus Chamfer Similarity to form a frame-to-frame similarity matrix; a four-layer CNN then reads that matrix and Chamfer Similarity reduces it to a video-to-video score. Because no global pooling collapses the frame or video into a vector before matching, the network can exploit diagonal and block-like temporal structures that indicate shared segments, and suppress isolated spurious frame matches. The paper supports the claim with ablations showing each component adds mAP on FIVR-5K, and with comparisons on CC_WEB_VIDEO, FIVR-200K, EVVE, and ActivityNet.

Load-bearing premise

The load-bearing premise is that the evaluation benchmarks' ground-truth labels are trustworthy enough that reported gains reflect true retrieval quality; the paper itself notes that CC_WEB_VIDEO contains many positives mislabeled as negatives and shows unlabeled related videos in FIVR-200K.

Editorial extensions

If this is right

  • On fine-grained incident retrieval (FIVR-200K), ViSiLv reports mAP of 0.892 on duplicate-scene retrieval, higher than the compared dynamic-programming and temporal-network reimplementations (0.775 and 0.724).
  • On activity retrieval with 3D convolutional features, the symmetric variant ViSiLsym reaches 0.745 mAP, outperforming the deep metric learning baseline's 0.705.
  • On the cleaned near-duplicate benchmark, ViSiLv approaches perfect ranking (0.993 mAP over the whole dataset), implying that the remaining errors are few once label noise is removed.
  • Because the method never collapses frames or videos into a single descriptor before matching, it can report a video-to-video score while retaining which segments matched, which is directly useful for localization-style retrieval tasks.

Reading between the lines

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

  • Inference: the two-stage design is a template: any region-level or spatio-temporal frame feature could be dropped into the TD+CS front end, so the method's gains may transfer to other backbones or modalities beyond the two feature types tested.
  • Inference: the asymmetric Chamfer Similarity outperforming its symmetric variant suggests that retrieval direction matters—query-to-database matching may benefit from treating the query as the reference—an effect worth testing on ranking fairness and duplicate-direction analysis.
  • Inference: the learned CNN on the similarity matrix could be applied to partial-copy detection and video re-localization directly by scanning submatrices of the frame-to-frame matrix, a direction the paper lists as future work.
  • Inference: the dependence on cleaned labels means a shared, independently verified annotation protocol across near-duplicate benchmarks would be needed to confirm the margin over prior methods.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper introduces ViSiL, a video similarity learning architecture that computes frame-to-frame similarity from regional CNN features via Tensor Dot and Chamfer Similarity, and then learns a video-to-video similarity score by passing the resulting frame-similarity matrix through a small CNN. The network is trained with a triplet loss on the VCDB dataset, using whitened and attention-weighted features, and is evaluated on four video retrieval tasks: near-duplicate video retrieval (CC_WEB_VIDEO), fine-grained incident video retrieval (FIVR-200K), event video retrieval (EVVE), and action video retrieval (ActivityNet). The authors report consistent improvements over the compared baselines and release the implementation publicly.

Significance. If the reported results hold, the paper makes a useful contribution by demonstrating that fine-grained spatial and temporal relations, captured through regional frame descriptors and a learned interaction on the frame-similarity matrix, are beneficial for video retrieval beyond global video embeddings. The method is well motivated and the design is technically sound in its components. The public release of the implementation is a concrete strength. However, the strength of the empirical claim is weakened by two evaluation issues that affect the interpretation of the state-of-the-art comparisons.

major comments (2)
  1. [§6.3.3, Table 7] The EVVE comparison is not a valid head-to-head comparison with LAMV. The paper states that ViSiL results are reported on roughly 80% of the EVVE dataset (1,897 of 2,375 videos and 503 of 620 queries), while the LAMV baseline numbers (0.536 and 0.587) are from the full dataset. Mean Average Precision is not invariant to subsetting the database and queries: the set of ground-truth positives per query changes, the ranking changes, and per-event mAP is computed over different query sets. Therefore the reported difference between ViSiLv (0.631, subset) and LAMV+QE (0.587, full) does not establish that ViSiL outperforms LAMV on EVR. The per-event comparisons in the same table suffer from the same problem. To support the stated 'large improvements' claim on EVR, the authors should either evaluate LAMV on the exact same subset, report results on the full EVVE dataset, or explicitly retract the state-of-the-art claim for EVR and present the results only as indicative on the available subset.
  2. [§5.1, Table 5] The near-perfect results on CC_WEB_VIDEO are reported on a 'cleaned' version of the annotations that the authors created, in addition to the original annotations. Because the cleaning is a post hoc modification of a public benchmark's ground truth, the numbers on the cleaned version are not directly comparable with previously published results on the original dataset. The paper does report original-annotation results as well, which is good, but the abstract's claim of 'large improvements' on NDVR relies substantially on the cleaned-annotation results (e.g., 0.996 vs. 0.985 on CC_WEBVIDEO_c). The authors should make the cleaning procedure and the list of changed labels publicly available and should clearly separate the two evaluations in all claims, rather than presenting the cleaned results as the primary evidence of superiority.
minor comments (5)
  1. [Section 2] The text refers to 'Goa et al.'; the correct name is Gao et al. (the reference is correctly listed as [11] and [35] in the bibliography).
  2. [Section 4.2, Eq. (4)] The notation d^T_{ij} b_{kl} is clear, but it may help to explicitly state that d and b are unit-norm region vectors so that the dot product is a cosine similarity, which is assumed in the attention weighting of Section 4.1.
  3. [Figure 2] The caption says 'The double arrows indicate shared weights,' but the figure contains only single arrows in the provided text version; please ensure the figure and caption are consistent.
  4. [Section 6.3.3, Table 7] The per-event class columns in Table 7 are presented without event names, only an ordering note referring to the original EVVE paper. It would improve reproducibility to print the event names in the table or in a supplementary file.
  5. [Section 5.2] The paper reports no variance or multiple-seed statistics for the training procedure. Since model selection is performed on a validation set (with mAP), reporting the mean and standard deviation over several training runs would strengthen the reliability of the comparisons, especially where margins are thin (e.g., ViSiLv vs. ViSiLsym on ActivityNet).

Circularity Check

0 steps flagged · score 2.0 of 10

No derivation-level circularity: ViSiL is trained on VCDB and evaluated on external benchmarks, but the EVVE comparison is benchmark-mismatched and the CC_WEB annotations are manually cleaned; these are evaluation-validity issues, not circular reductions.

full rationale

ViSiL's derivation chain is self-contained: frame features are extracted by a fixed CNN with PCA whitening learned from VCDB, frame similarity is Chamfer Similarity on Tensor Dot products of regional features, video similarity is the output of a CNN trained with triplet and regularization losses, and the trained model is evaluated on CC_WEB_VIDEO, FIVR-200K, EVVE, and ActivityNet. No test-set labels enter training, and no fitted parameter is renamed as a prediction. The attention vector u and the video-similarity CNN weights are learned from Equations 7-9, not set to reproduce benchmark scores. The use of the authors' prior BoW method [20] for triplet mining and for defining FIVR-5K is a training/subset-construction choice, not a reduction of the reported mAP values. The main caveats are evaluation-validity issues rather than circularity: (i) the EVVE comparison in Table 7 is reported on about 80% of the dataset while the LAMV baselines are from the full dataset, so the claimed EVR improvement is not established by that table; (ii) CC_WEB_VIDEO results are reported on the authors' manually 'cleaned' annotations, so near-perfect scores on the cleaned labels are not fully independent of that relabeling. Neither issue makes the architecture's output equivalent by definition to its training input or to a self-citation. The central result has independent grounding in the FIVR-200K and ActivityNet evaluations, and the self-citations (DML, LBoW, FIVR dataset) are not load-bearing in the derivation.

Assumptions & free parameters 5 free parameters · 4 assumptions · 1 invented entities

The method rests on standard supervised learning with validation-set hyperparameter selection. The listed free parameters are the explicit knobs that affect the reported mAP. The axioms capture the domain assumptions about feature quality, the Chamfer design choice, training distribution match, and the bounded-output regularization mechanism.

free parameters (5)
  • triplet margin gamma = 0.5
    Trip let loss margin; selected by validation. Supplementary Figure 5(a) shows sensitivity: larger gamma helps DSVR but hurts ISVR.
  • regularization weight r = 0.1
    Weight of the similarity regularization loss; selected by validation. Supplementary Figure 5(b) shows r=1.0 improves DSVR/CSVR but default 0.1 is better for ISVR.
  • training snippet size W = 64
    Number of frames sampled per video during training; selected by validation. Larger W helps DSVR/CSVR.
  • region granularity N = 3 (L3-iMAC)
    Number of regions per frame dimension; N=3 selected after FIVR-5K experiments in Table 2.
  • PCA whitening dimensionality = 256 (L3-iMAC)
    Dimensionality of region vectors after PCA; 9x256 configuration chosen as a trade-off in Table 2.
assumptions (4)
  • domain assumption Pretrained CNN features (ResNet-50, I3D) provide representations that preserve enough visual information for similarity comparison.
    Feature extraction in Section 4.1 and implementation in Section 5.2 rely on fixed pretrained CNNs.
  • domain assumption Chamfer similarity over regional features is an appropriate frame-to-frame similarity function, with max-pooling over regions giving spatial invariance.
    The entire frame and video similarity computation in Section 4.2 and 4.3 is built on this choice; no proof is given.
  • domain assumption The VCDB training set and artificially transformed videos provide a training distribution that generalizes to the evaluation tasks (incident, event, action retrieval).
    Training in Section 4.5 and 5.1 uses VCDB only, yet the method is applied to four different tasks.
  • ad hoc to paper Hard tanh clipping plus a regularization loss keeps outputs bounded and improves training.
    Section 4.4 introduces this mechanism without a formal justification.
invented entities (1)
  • Visual context vector u
    purpose: Learned saliency weighting for frame regions in the attention mechanism.
    The vector is a learned parameter inside the network (Section 4.1); it is not verified outside the model and no external prediction is made from it.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ViSiL: Fine-grained Spatio-Temporal Video Similarity Learning." pith.science (2026). https://pith.science/paper/GH4QI74V

@misc{pith2026190807410,
  author       = {Pith},
  title        = {Pith review of: ViSiL: Fine-grained Spatio-Temporal Video Similarity Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GH4QI74V}},
  note         = {Machine review of arXiv:1908.07410}
}
read the original abstract

In this paper we introduce ViSiL, a Video Similarity Learning architecture that considers fine-grained Spatio-Temporal relations between pairs of videos -- such relations are typically lost in previous video retrieval approaches that embed the whole frame or even the whole video into a vector descriptor before the similarity estimation. By contrast, our Convolutional Neural Network (CNN)-based approach is trained to calculate video-to-video similarity from refined frame-to-frame similarity matrices, so as to consider both intra- and inter-frame relations. In the proposed method, pairwise frame similarity is estimated by applying Tensor Dot (TD) followed by Chamfer Similarity (CS) on regional CNN frame features - this avoids feature aggregation before the similarity calculation between frames. Subsequently, the similarity matrix between all video frames is fed to a four-layer CNN, and then summarized using Chamfer Similarity (CS) into a video-to-video similarity score -- this avoids feature aggregation before the similarity calculation between videos and captures the temporal similarity patterns between matching frame sequences. We train the proposed network using a triplet loss scheme and evaluate it on five public benchmark datasets on four different video retrieval problems where we demonstrate large improvements in comparison to the state of the art. The implementation of ViSiL is publicly available.

Figures

Figures reproduced from arXiv: 1908.07410 by the authors.

Figure 1
Figure 1. Depiction of the frame-to-frame similarity matrix and the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the training scheme of the proposed architecture. A triplet of an anchor, positive and negative videos is provided to [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Illustration of frame-level similarity calculation between [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Examples of challenging cases of related videos that [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Impact of the margin hyperparameter γ, the regularization parameter r and video snippet size W on the performance of the proposed method on FIVR-5K. Run MAC iMAC ViSiLf ViSiLv Comp. Time Offline Online 0.95s 2.0ms 0.95s 2.3ms 0.96s 6.0ms 1.04s 9.5ms FIVR-5K DSVR CSVR I…
Figure 6
Figure 6. Figure 6: Examples of the attention weighting on arbitrary video frames: sampled video frames from the same video (top), attention maps [PITH_FULL_IMAGE:figures/full_fig_p013_6.png]
Figure 7
Figure 7. Figure 7: Visual examples of the input and output of ViSiL for three different video relation types. Two sampled frames of the compared [PITH_FULL_IMAGE:figures/full_fig_p013_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 34 canonical work pages

  1. [1]

    Aggregating local deep features for image retrieval

    Artem Babenko and Victor Lempitsky. Aggregating local deep features for image retrieval. InProceedings of the IEEE International Conference on Computer Vision, pages 1269– 1277, 2015

    work page 2015

  2. [2]

    LAMV: Learning to align and match videos with ker- nelized temporal layers

    Lorenzo Baraldi, Matthijs Douze, Rita Cucchiara, and Herv ´e J´egou. LAMV: Learning to align and match videos with ker- nelized temporal layers. In Proceedings of the IEEE con- ference on Computer Vision and Pattern Recognition, pages 7804–7813, 2018

    work page 2018

  3. [3]

    Parametric correspondence and chamfer matching: Two new techniques for image matching

    Harry G Barrow, Jay M Tenenbaum, Robert C Bolles, and Helen C Wolf. Parametric correspondence and chamfer matching: Two new techniques for image matching. Techni- cal report, SRI AI Center, 1977

    work page 1977

  4. [4]

    Activitynet: A large-scale video benchmark for human activity understanding

    Fabian Caba Heilbron, Victor Escorcia, Bernard Ghanem, and Juan Carlos Niebles. Activitynet: A large-scale video benchmark for human activity understanding. In Proceed- ings of the IEEE conference on Computer Vision and Pattern Recognition, pages 961–970, 2015

    work page 2015

  5. [5]

    Million-scale near-duplicate video retrieval system

    Yang Cai, Linjun Yang, Wei Ping, Fei Wang, Tao Mei, Xian- Sheng Hua, and Shipeng Li. Million-scale near-duplicate video retrieval system. In Proceedings of the 19th ACM international conference on Multimedia , pages 837–838. ACM, 2011

    work page 2011

  6. [6]

    Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset

    Joao Carreira and Andrew Zisserman. Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, pages 4724–4733. IEEE, 2017

    work page 2017

  7. [7]

    Pattern- based near-duplicate video retrieval and localization on web- scale videos

    Chien-Li Chou, Hua-Tsung Chen, and Suh-Yin Lee. Pattern- based near-duplicate video retrieval and localization on web- scale videos. IEEE Transactions on Multimedia, 17(3):382– 395, 2015

    work page 2015

  8. [8]

    An image-based approach to video copy detection with spatio- temporal post-filtering

    Matthijs Douze, Herv ´e J ´egou, and Cordelia Schmid. An image-based approach to video copy detection with spatio- temporal post-filtering. IEEE Transactions on Multimedia , 12(4):257–266, 2010

    work page 2010

Show all 37 references
  1. [9]

    Stable hyper-pooling and query expansion for event detection

    Matthijs Douze, J ´erˆome Revaud, Cordelia Schmid, and Herv´e J´egou. Stable hyper-pooling and query expansion for event detection. In Proceedings of the IEEE International Conference on Computer Vision, pages 1825–1832, 2013

    2013

  2. [10]

    Video re-localization

    Yang Feng, Lin Ma, Wei Liu, Tong Zhang, and Jiebo Luo. Video re-localization. In Proceedings of the European Con- ference on Computer Vision, pages 51–66, 2018

    2018

  3. [11]

    ER3: A unified framework for event retrieval, recognition and recounting

    Zhanning Gao, Gang Hua, Dongqing Zhang, Nebojsa Jojic, Le Wang, Jianru Xue, and Nanning Zheng. ER3: A unified framework for event retrieval, recognition and recounting. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition, pages 2253–2262, 2017

    2017

  4. [12]

    Unsupervised t-distributed video hashing and its deep hashing extension

    Yanbin Hao, Tingting Mu, John Y Goulermas, Jianguo Jiang, Richang Hong, and Meng Wang. Unsupervised t-distributed video hashing and its deep hashing extension. IEEE Trans- actions on Image Processing, 26(11):5531–5544, 2017

    2017

  5. [13]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceed- ings of the IEEE conference on Computer Vision and Pattern Recognition, pages 770–778, 2016

    2016

  6. [14]

    Learning spatial-temporal fea- tures for video copy detection by the combination of cnn and rnn

    Yaocong Hu and Xiaobo Lu. Learning spatial-temporal fea- tures for video copy detection by the combination of cnn and rnn. Journal of Visual Communication and Image Represen- tation, 55:21–29, 2018

    2018

  7. [15]

    Negative evidences and co- occurences in image retrieval: The benefit of pca and whiten- ing

    Herv ´e J´egou and Ondˇrej Chum. Negative evidences and co- occurences in image retrieval: The benefit of pca and whiten- ing. In Proceedings of the European Conference on Com- puter Vision, pages 774–787. Springer, 2012

    2012

  8. [16]

    VCDB: a large-scale database for partial copy detection in videos

    Yu-Gang Jiang, Yudong Jiang, and Jiajun Wang. VCDB: a large-scale database for partial copy detection in videos. In Proceedings of the European Conference on Computer Vi- sion, pages 357–371. Springer, 2014

    2014

  9. [17]

    Partial copy detection in videos: A benchmark and an evaluation of popular methods

    Yu-Gang Jiang and Jiajun Wang. Partial copy detection in videos: A benchmark and an evaluation of popular methods. IEEE Transactions on Big Data, 2(1):32–42, 2016

    2016

  10. [18]

    Adam: A method for stochastic optimization

    Diederik P Kingma and Jimmy Ba. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980 , 2014

    2014 arXiv

  11. [19]

    FIVR: Fine-grained In- cident Video Retrieval

    Giorgos Kordopatis-Zilos, Symeon Papadopoulos, Ioannis Patras, and Ioannis Kompatsiaris. FIVR: Fine-grained In- cident Video Retrieval. arXiv preprint arXiv:1809.04094 , 2018

    2018 arXiv

  12. [20]

    Near-duplicate video re- trieval by aggregating intermediate cnn layers

    Giorgos Kordopatis-Zilos, Symeon Papadopoulos, Ioannis Patras, and Yiannis Kompatsiaris. Near-duplicate video re- trieval by aggregating intermediate cnn layers. In Interna- tional conference on Multimedia Modeling, pages 251–263. Springer, 2017

    2017

  13. [21]

    Near-duplicate video re- trieval with deep metric learning

    Giorgos Kordopatis-Zilos, Symeon Papadopoulos, Ioannis Patras, and Yiannis Kompatsiaris. Near-duplicate video re- trieval with deep metric learning. InProceedings of the IEEE International Conference on Computer Vision , pages 347– 356, 2017

    2017

  14. [22]

    IR feature embedded bof indexing method for near-duplicate video re- trieval

    Kaiyang Liao, Hao Lei, Yuanlin Zheng, Guangfeng Lin, Congjun Cao, Mingzhu Zhang, and Jie Ding. IR feature embedded bof indexing method for near-duplicate video re- trieval. IEEE Transactions on Circuits and Systems for Video Technology, 2018

    2018

  15. [23]

    Deep video hashing

    Venice Erin Liong, Jiwen Lu, Yap-Peng Tan, and Jie Zhou. Deep video hashing. IEEE Transactions on Multimedia , 19(6):1209–1219, 2017

    2017

  16. [24]

    An image-based near-duplicate video retrieval and localization using improved edit distance

    Hao Liu, Qingjie Zhao, Hao Wang, Peng Lv, and Yanming Chen. An image-based near-duplicate video retrieval and localization using improved edit distance. Multimedia Tools and Applications, 76(22):24435–24456, 2017

    2017

  17. [25]

    Near-duplicate video re- trieval: Current research and future trends

    Jiajun Liu, Zi Huang, Hongyun Cai, Heng Tao Shen, Chong Wah Ngo, and Wei Wang. Near-duplicate video re- trieval: Current research and future trends. ACM Computing Surveys (CSUR), 45(4):44, 2013

    2013

  18. [26]

    Temporal matching kernel with explicit feature maps

    S ´ebastien Poullot, Shunsuke Tsukatani, Anh Phuong Nguyen, Herv ´e J ´egou, and Shin’Ichi Satoh. Temporal matching kernel with explicit feature maps. In Proceedings of the 23rd ACM international conference on Multimedia, pages 381–390. ACM, 2015

    2015

  19. [27]

    Fine- tuning CNN image retrieval with no human annotation.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018

    Filip Radenovi ´c, Giorgos Tolias, and Ondrej Chum. Fine- tuning CNN image retrieval with no human annotation.IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018

    2018

  20. [28]

    Event retrieval in large video collections with circulant temporal encoding

    J ´erˆome Revaud, Matthijs Douze, Cordelia Schmid, and Herv´e J´egou. Event retrieval in large video collections with circulant temporal encoding. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 2459–2466, 2013

    2013

  21. [29]

    Very deep convo- lutional networks for large-scale image recognition

    Karen Simonyan and Andrew Zisserman. Very deep convo- lutional networks for large-scale image recognition. arXiv preprint arXiv:1409.1556, 2014

    2014 arXiv

  22. [30]

    Multiple feature hashing for real-time large scale near-duplicate video retrieval

    Jingkuan Song, Yi Yang, Zi Huang, Heng Tao Shen, and Richang Hong. Multiple feature hashing for real-time large scale near-duplicate video retrieval. In Proceedings of the 19th ACM international conference on Multimedia , pages 423–432. ACM, 2011

    2011

  23. [31]

    Self-supervised video hashing with hierarchical binary auto-encoder

    Jingkuan Song, Hanwang Zhang, Xiangpeng Li, Lianli Gao, Meng Wang, and Richang Hong. Self-supervised video hashing with hierarchical binary auto-encoder. IEEE Trans- actions on Image Processing, 27(7):3210–3221, 2018

    2018

  24. [32]

    Scalable detection of partial near-duplicate videos by visual-temporal consistency

    Hung-Khoon Tan, Chong-Wah Ngo, Richard Hong, and Tat- Seng Chua. Scalable detection of partial near-duplicate videos by visual-temporal consistency. InProceedings of the 17th ACM international conference on Multimedia , pages 145–154. ACM, 2009

    2009

  25. [33]

    Particular object retrieval with integral max-pooling of cnn activations

    Giorgos Tolias, Ronan Sicre, and Herv ´e J ´egou. Particular object retrieval with integral max-pooling of cnn activations. arXiv preprint arXiv:1511.05879, 2015

    2015 arXiv

  26. [34]

    Learning spatiotemporal features with 3d convolutional networks

    Du Tran, Lubomir Bourdev, Rob Fergus, Lorenzo Torresani, and Manohar Paluri. Learning spatiotemporal features with 3d convolutional networks. InProceedings of the IEEE Inter- national Conference on Computer Vision, pages 4489–4497, 2015

    2015

  27. [35]

    Practical elimination of near-duplicates from web video search

    Xiao Wu, Alexander G Hauptmann, and Chong-Wah Ngo. Practical elimination of near-duplicates from web video search. In Proceedings of the 15th ACM international con- ference on Multimedia, pages 218–227. ACM, 2007

    2007

  28. [36]

    Deep multi-task representation learning: A tensor factorisation approach

    Yongxin Yang and Timothy Hospedales. Deep multi-task representation learning: A tensor factorisation approach. In International Conference on Learning Representations , 2017

    2017

  29. [37]

    Hierarchical attention networks for document classification

    Zichao Yang, Diyi Yang, Chris Dyer, Xiaodong He, Alex Smola, and Eduard Hovy. Hierarchical attention networks for document classification. In Proceedings of the 2016 Con- ference of the North American Chapter of the Association for Computational Linguistics: Human Language Tech...

    2016

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.