http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-112883896-B
Outgoing Links
Predicate | Object |
---|---|
classificationCPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-045 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06V40-168 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-08 http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06V40-174 |
classificationIPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06N3-04 http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06N3-08 http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06V10-82 http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06V20-40 http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06V40-16 |
filingDate | 2021-03-10-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
grantDate | 2022-10-11-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
publicationDate | 2022-10-11-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
publicationNumber | CN-112883896-B |
titleOfInvention | A method of micro-expression detection based on BERT network |
abstract | The invention relates to a micro-expression detection method based on a BERT network, comprising: A. preprocessing a micro-expression video; B. dividing an optical flow sequence into a training set and a test set, and calculating the intersection ratio IoU as the mark of the segment value; C. Build a spatiotemporal feature extraction model, including a three-dimensional convolutional neural network and a BERT network; D. Train a spatiotemporal feature extraction model; E. Evaluate the intersection ratio of candidate micro-expression segments in the test set based on the trained spatiotemporal feature extraction model IoU; F. Grouping and merging consecutive candidate micro-expression segments with high IoU values, and suppressing overlapping segments to form the final micro-expression detection result. The present invention extracts the features in each time slot, and in order to better pay attention to the importance of different time slots in the segment, the attention mechanism of BERT is used to extract the features between different time slots, and the information of the whole video sequence is fully utilized. |
priorityDate | 2021-03-10-04:00^^<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/compound/CID6442267 http://rdf.ncbi.nlm.nih.gov/pubchem/substance/SID426109005 |
Total number of triples: 19.