http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-109299341-B
Outgoing Links
Predicate | Object |
---|---|
classificationCPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06F18-24147 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06F18-2413 |
classificationIPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06F16-903 http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06K9-62 |
filingDate | 2018-10-29-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
grantDate | 2020-05-05-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
publicationDate | 2020-05-05-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
publicationNumber | CN-109299341-B |
titleOfInvention | An adversarial cross-modal retrieval method and system based on dictionary learning |
abstract | The invention discloses a method and system for confrontational cross-modal retrieval based on dictionary learning. The method includes: acquiring underlying features of image data and text data, and constructing training sets and test sets of images and texts based on the underlying features; constructing The dictionary learning model is trained based on the training set of images and texts, and according to the obtained image dictionary and text dictionary, new training sets and test sets are constructed; the new training sets of images and texts are projected to a common representation space; To jointly represent the image and text feature data in the space, learn the feature holder, that is, to perform feature discrimination and triple sorting, and learn the modal classifier; perform adversarial learning on the feature holder and the modal classifier to optimize the common representation space, using Test set for cross-modal retrieval. Using dictionary learning for feature extraction and adversarial learning to better learn the common space of image modalities and text modalities can greatly improve the accuracy of cross-modal retrieval. |
priorityDate | 2018-10-29-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
type | http://data.epo.org/linked-data/def/patent/Publication |
Incoming Links
Predicate | Subject |
---|---|
isDiscussedBy | http://rdf.ncbi.nlm.nih.gov/pubchem/substance/SID411371601 http://rdf.ncbi.nlm.nih.gov/pubchem/compound/CID40521 |
Total number of triples: 14.