http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-110263666-A

Outgoing Links

Predicate Object
assignee http://rdf.ncbi.nlm.nih.gov/pubchem/patentassignee/MD5_eda37409f26cb8e833ba67f2b1bc1fd5
classificationCPCInventive http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-045
http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06V20-41
http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06V20-49
http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06N3-08
classificationIPCInventive http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06N3-04
http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06K9-00
http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06N3-08
filingDate 2019-05-29-04:00^^<http://www.w3.org/2001/XMLSchema#date>
inventor http://rdf.ncbi.nlm.nih.gov/pubchem/patentinventor/MD5_0ff6a5d4fcf5706e17b088876baa2114
http://rdf.ncbi.nlm.nih.gov/pubchem/patentinventor/MD5_c32f42b1a434524927fc2c31d4bf430b
http://rdf.ncbi.nlm.nih.gov/pubchem/patentinventor/MD5_4b7bb4a04d53ab06bbec605fed6328c4
http://rdf.ncbi.nlm.nih.gov/pubchem/patentinventor/MD5_ec4735c4906eb783c0791b5f33b863ba
publicationDate 2019-09-20-04:00^^<http://www.w3.org/2001/XMLSchema#date>
publicationNumber CN-110263666-A
titleOfInvention An Action Detection Method Based on Asymmetric Multi-Stream
abstract The invention discloses an action detection method based on asymmetric multi-stream, comprising the following steps: extracting RGB image and optical flow from prior video, training to obtain trained RGB image single-stream network and optical flow single-stream network; extracting prior The image flow feature information and optical flow feature information of each frame in the video, combined with the action label, trains an asymmetric two-stream network; through the trained RGB image single-stream network and optical flow single-stream network, extract the target video of each frame to be detected. Image flow feature information and optical flow feature information, obtain the segment features of the target video and input it into the trained asymmetric two-stream network, and calculate and obtain the video classification vector; select potential actions from the video classification vector, and obtain the action recognition sequence of potential actions; through The action recognition sequence completes the detection of actions. The motion detection method of the present invention takes into account the asymmetry between image flow and optical flow, and can improve the accuracy of motion recognition and motion detection.
isCitedBy http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-113836969-A
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-111709410-A
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-111914644-A
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-111709410-B
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-112800941-A
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-112464856-A
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-110942037-A
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-112464856-B
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/WO-2022121543-A1
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-110866938-A
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-113298013-A
http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-110866938-B
priorityDate 2019-05-29-04:00^^<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/SID415713197
http://rdf.ncbi.nlm.nih.gov/pubchem/compound/CID25572

Total number of triples: 33.