Improving the search accuracy of the VLAD through weighted aggregation of local descriptors

Kim, Tak-Eun; Kim, Myoung Ho

doi:10.1016/j.jvcir.2015.07.005

Cited by 11 publications

(3 citation statements)

References 41 publications

(72 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…where T represents the total number of frames under consideration. Kim and Kim [22] propose an extension to the original VLAD approach by weighting the descriptors depending on their importance. For determining the spatial image locations which contain discriminant information, a saliency map of the image is used.…”

Section: Vlad Encodingmentioning

confidence: 99%

Top-down attention recurrent VLAD encoding for action recognition in videos

Sudhakaran

Lanz

2019

View full text Add to dashboard Cite

Most recent approaches for action recognition from video leverage deep architectures to encode the video clip into a fixed length representation vector that is then used for classification. For this to be successful, the network must be capable of suppressing irrelevant scene background and extract the representation from the most discriminative part of the video. Our contribution builds on the observation that spatiotemporal patterns characterizing actions in videos are highly correlated with objects and their location in the video. We propose Top-down Attention Action VLAD (TA-VLAD), a deep recurrent architecture with built-in spatial attention that performs temporally aggregated VLAD encoding for action recognition from videos. We adopt a top-down approach of attention, by using class specific activation maps obtained from a deep CNN pre-trained for image classification, to weight appearance features before encoding them into a fixed-length video descriptor using Gated Recurrent Units. Our method achieves state of the art recognition accuracy on HMDB51 and UCF101 benchmarks.

show abstract

Section: Vlad Encodingmentioning

confidence: 99%

Top-down attention recurrent VLAD encoding for action recognition in videos

Sudhakaran

Lanz

2019

View full text Add to dashboard Cite

show abstract

“…Based on these pioneering works, aggregated vector-based methods emerged, including vector quantization [3], sparse coding [4], localityconstrained linear coding [5], and soft assignment [6]. Aggregated vector-based encoding methods succeed in various image retrieval applications [3][4][5][7][8][9]. The vector of locally aggregated descriptors (VLAD) [8] is one of the most widely adopted aggregated vector-based methods.…”

Section: Introductionmentioning

confidence: 99%

Adding spatial distribution clue to aggregated vector in image retrieval

Zhang

Guo

et al. 2018

J Image Video Proc.

View full text Add to dashboard Cite

This study proposes a novel algorithm that enhances the distinctiveness of the traditional vector of locally aggregated descriptors (VLAD) using spatial distribution clue of local features. The algorithm introduces a new method to compute the spatial distribution entropy (SDE) of clusters. Unlike conventional methods, this algorithm considers the distribution of full spatial information provided by local feature detectors rather than only utilizing the spatial coordinate statistics. For each cluster, the corresponding spatial distribution is computed using a histogram of spatial locations, scales, and orientations of all local features inside the cluster. Entropy is calculated from the spatial distributions of all clusters of an image to create a distribution function, which is further normalized and concatenated with the VLAD vector to generate the final representation. Image retrieval and classification experiments on public datasets are performed. Experimental results show that the proposed algorithms produce better or comparable retrieval performance than several state-of-the-art algorithms. In addition, we extend our SDE to the convolutional neural network (CNN) feature, which further improves the CNN feature result in image retrieval.

show abstract

“…BoW builds a high-dimensional sparse histogram as a global feature for an image. There are three reasons for the success of BoW [9] representations: they are based on local invariant features, they can be compared with standard distances, and they can rely on an inverted list to boost their retrieval efficiency. Nevertheless, BoW has some drawbacks [10].…”

Section: Introductionmentioning

confidence: 99%

Distribution Entropy Boosted VLAD for Image Retrieval

Zhou

Wang

et al. 2016

Entropy

View full text Add to dashboard Cite

Several recent works have shown that aggregating local descriptors to generate global image representation results in great efficiency for retrieval and classification tasks. The most popular method following this approach is VLAD (Vector of Locally Aggregated Descriptors). We present a novel image presentation called Distribution Entropy Boosted VLAD (EVLAD), which extends the original vector of locally aggregated descriptors. The original VLAD adopts only residuals to depict the distribution information of every visual word and neglects other statistical clues, so its discriminative power is limited. To address this issue, this paper proposes the use of the distribution entropy of each cluster as supplementary information to enhance the search accuracy. To fuse two feature sources organically, two fusion methods after a new normalization stage meeting power law are also investigated, which generate identically sized and double-sized vectors as the original VLAD. We validate our approach in image retrieval and image classification experiments. Experimental results demonstrate the effectiveness of our algorithm.

show abstract

Improving the search accuracy of the VLAD through weighted aggregation of local descriptors

Cited by 11 publications

References 41 publications

Top-down attention recurrent VLAD encoding for action recognition in videos

Top-down attention recurrent VLAD encoding for action recognition in videos

Adding spatial distribution clue to aggregated vector in image retrieval

Distribution Entropy Boosted VLAD for Image Retrieval

Contact Info

Product

Resources

About