Adding spatial distribution clue to aggregated vector in image retrieval

Li, Pingping; Zhang, Miao; Guo, Huili; Wang, Ye‐Ran; Ni, Ai

doi:10.1186/s13640-018-0247-0

Cited by 12 publications

(8 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Sanchezet [41] added the spatial position information of features into the descriptor, which overcame the change in the proportion of retrieved objects and the change in the local area of the image. Liu [42] introduced the concept of spatial distribution entropy and combined it with the VLAD algorithm. These methods have fully introduced the spatial information into the image retrieval algorithm, and great effects are obtained.…”

Section: Application Of Spatial Information In Image Retrievalmentioning

confidence: 99%

Relative Distribution Entropy Loss Function in CNN Image Retrieval

Shi

Zhang

et al. 2020

Entropy

Self Cite

View full text Add to dashboard Cite

Convolutional neural networks (CNN) is the most mainstream solution in the field of image retrieval. Deep metric learning is introduced into the field of image retrieval, focusing on the construction of pair-based loss function. However, most pair-based loss functions of metric learning merely take common vector similarity (such as Euclidean distance) of the final image descriptors into consideration, while neglecting other distribution characters of these descriptors. In this work, we propose relative distribution entropy (RDE) to describe the internal distribution attributes of image descriptors. We combine relative distribution entropy with the Euclidean distance to obtain the relative distribution entropy weighted distance (RDE-distance). Moreover, the RDE-distance is fused with the contrastive loss and triplet loss to build the relative distributed entropy loss functions. The experimental results demonstrate that our method attains the state-of-the-art performance on most image retrieval benchmarks.

show abstract

Section: Application Of Spatial Information In Image Retrievalmentioning

confidence: 99%

Relative Distribution Entropy Loss Function in CNN Image Retrieval

Shi

Zhang

et al. 2020

Entropy

Self Cite

View full text Add to dashboard Cite

show abstract

“…′gj cx = (gj cx -di cx )/di w ′gj cy = (gj cy -di cy )/di h (4) ′gj w = log(gj w / di w ) ′gj h = log(gj h / di h ) (5) In which,(g cx , g cy , g w , g h )represents the fact library box,(d cx , d cy , d w , d h )indicates the default box, (l cx , l cy , l w , l h )indicates the offset of the predicted box relative to the default box.…”

Section: Algorithm Principlementioning

confidence: 99%

“…Therefore, the description of the station logo is the first and most critical step in the station logo detection. At present, the existing domestic standard feature analysis algorithms are: based on color histogram [1], ordinary Hu invariant moment [2], weighted Hu invariant moment [3], spatial distribution histogram [4] and so on. The TV station logo detection based on color histogram uses different color tones between different types of station labels to complete the station caption detection.…”

Section: Introductionmentioning

confidence: 99%

A Detection algorithm based on Convolutional Neural Network

Fei¹,

Li²,

Xu³

et al. 2018

Preprint

View full text Add to dashboard Cite

The Station logo is a way for a TV station to claim copyright, which can realize the analysis and understanding of the video by the identification of the station logo, so as to ensure that the broadcasted TV signal will not be illegally interfered. In this paper, we design a station logo detection method based on Convolutional Neural Network by the characteristics of the station, such as small scale-to-height ratio change and relatively fixed position. Firstly, in order to realize the preprocessing and feature extraction of the station data, the video samples are collected, filtered, framed, labeled and processed. Then, the training sample data and the test sample data are divided proportionally to train the station detection model. Finally, the sample is tested to evaluate the effect of the training model in practice. The simulation experiments prove its validity.

show abstract

“…Another way to obtain similarity in the images is that applied in [6] where the proposed algorithm improves distinctive characters of the image using spa-tial distribution, which is calculated by means of the location histogram. On space, scales and orientations of the local characteristics, in order to achieve a good recovery.…”

Section: Introductionmentioning

confidence: 99%

Molecular Marker for Protein Electrophoresis by Wavelet Transform

Flores-López¹,

Flores-Pulido²,

Jaramillo-Quintero³

2019

RCS

View full text Add to dashboard Cite

Visual Information retrieval is an area where the analysis and recognition is helpful to generate and obtain data from image samples. In this research the image samples are obtained from analysis of electrophoresis that captures protein profiles in tissues. The electrophoresis provides a sample where the proteins are computed to know their molecular weight. The image digital processing takes electrophoresis images and a wavelet transform is applied to the sample that can be fractioned to emphasize the molecular weights that at first sight are not identified. Therefore, with the wavelet transform, it is possible to compute molecular weights of proteins and know their corresponding weight. Electrophoresis is a technique that is used in various analysis such as DNA, medicine, environment and food. All these profiles have different molecular weights, which are known by a marker that is placed in a lane. In this proposal, the wavelet transform is applied to the images of electrophoresis samples, creating the signal of the protein with the approximation coefficients, that achieves to measure a molecular weight. The approximation coefficients are computed at 3 levels of decomposition with the wavelet transform Daubechies. It is then possible to detect a molecular weight of nearly 300 samples of electrophoresis with an accuracy of 97 % in terms of numerical and visual similarity by means of visual information retrieval evaluated with recall and precision metrics.

show abstract

Adding spatial distribution clue to aggregated vector in image retrieval

Cited by 12 publications

References 49 publications

Relative Distribution Entropy Loss Function in CNN Image Retrieval

Relative Distribution Entropy Loss Function in CNN Image Retrieval

A Detection algorithm based on Convolutional Neural Network

Molecular Marker for Protein Electrophoresis by Wavelet Transform

Contact Info

Product

Resources

About