Similar retrieval method, equipment and storage medium for massive feature vector data
A feature vector and similarity retrieval technology, applied in the field of unstructured data search, can solve the problems of low retrieval efficiency of massive feature data, inability to guarantee recall rate and precision rate, etc., and achieve the effect of solving low retrieval efficiency.
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
no. 1 example
[0024] Such as figure 1 As shown, the first embodiment of the present invention provides a similar retrieval method for massive feature vector data, the method comprising steps:
[0025] S11. According to the feature vectors to be retrieved, perform calculations to obtain the coarse classification identifier after the rough classification hash coding, the binary code set after the multi-index hash coding, and the feature vector set.
[0026] In this embodiment, the feature vectors to be retrieved are vectors extracted from unstructured data such as images, videos, and voices.
[0027] In one embodiment, the rough classification identifier (ID) after the rough classification hash code is calculated in the following manner:
[0028] Calculate the feature vector and log to be retrieved 2 The inner product of S rough classification hash functions, where S is the number of classification labels;
[0029] In this embodiment, the value range of S may be less than or equal to 16. ...
no. 2 example
[0078] refer to figure 2 , figure 2 A similar retrieval device for massive eigenvector data provided by the third embodiment of the present invention, the device includes: a memory 21, a processor 22, and a computer that is stored in the memory 21 and can run on the processor 22 A similar retrieval program for massive eigenvector data, when the similar retrieval program for massive eigenvector data is executed by the processor 22, it is used to realize the following steps of the similar retrieval method for massive eigenvector data:
[0079] According to the feature vector to be retrieved, calculate and obtain the coarse classification identifier after the rough classification hash coding, the binary code set after the multi-index hash coding, and the feature vector set;
[0080] performing a joint search according to the rough classification identifier and the binary code set to obtain a joint search result set;
[0081] The joint search result set is filtered layer by la...
no. 3 example
[0102] The third embodiment of the present invention provides a computer-readable storage medium, the computer-readable storage medium stores a similarity retrieval program for massive eigenvector data, and the similarity retrieval program for massive eigenvector data is implemented when executed by a processor. Steps of the method for similarity retrieval of massive feature vector data described in the first embodiment.
[0103] The computer-readable storage medium provided by the embodiments of the present invention performs similarity retrieval on massive feature vector data through rough classification identifiers, binary code sets, and feature vector sets; it solves the problem of low retrieval efficiency and inability to retrieve massive feature data in the prior art Guaranteed recall and precision issues.
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More