Since a long time ago, supermarkets have become people's destinations for shopping for various things such as food, cooking ingredients, cleaning products and others. Supermarkets are known for their very large and crowded places, making it difficult to monitor. Therefore, supermarkets need a system to help monitoring. With the development of technology, monitoring systems are increasingly advanced and one of the results of these technological developments is a system for recognizing human activities. By using OpenPose to obtain human skeleton data on the image and using the Long Short Term Memory method to perform recognition, testing of the training data was carried out so as to produce a precision value of 99%, recall 99%, and f1-score 99%. And real-time testing using a camera resulted in an accuracy value of 73% for the picking class, 87% for the standing class and 81% for the walking class.
Copyrights © 2022