An analytics device (200) for monitoring maintenance on an elevator system performed by an individual including: a processor (282); and a memory (284) including computer-executable instructions that, when executed by the processor (282), cause the processor (282) to perform operations, the operations including: capturing a first video stream (310) using a first video camera (490a-d); extracting sequences from at least the first video stream (310); extracting features from the sequences; and analyzing, using a long short-term memory model, the sequence to determine whether the maintenance performed on the elevator system by the individual is performed correctly; wherein the long short-term memory model is trained by: obtaining training videos of individuals performing maintenance activities; labeling the training videos; extracting frames and clips from the training videos; extracting sequences from the frames and clips; and extracting features from the sequences to produce the long short-term memory model.
MACHINE LEARNING BASED HUMAN ACTIVITY DETECTION AND CLASSIFICATION IN FIRST AND THIRD PERSON VIDEOS
AUF MASCHINELLEM LERNEN BASIERENDE DETEKTION UND KLASSIFIZIERUNG MENSCHLICHER AKTIVITÄTEN IN VIDEOS DER ERSTEN UND DRITTEN PERSON
DÉTECTION D'ACTIVITÉ HUMAINE BASÉE SUR L'APPRENTISSAGE AUTOMATIQUE ET CLASSIFICATION DANS DES VIDÉOS À LA PREMIÈRE ET À LA TROISIÈME PERSONNES
2024-03-27
Patent
Electronic Resource
English
IPC: | B66B Aufzüge , ELEVATORS |
MACHINE LEARNING BASED HUMAN ACTIVITY DETECTION AND CLASSIFICATION IN FIRST AND THIRD PERSON VIDEOS
European Patent Office | 2024
|DEEP INTERSECTION CLASSIFICATION USING FIRST AND THIRD PERSON VIEWS
British Library Conference Proceedings | 2019
|Incremental learning of human activity models from videos
British Library Online Contents | 2016
|