Crearea unei rețele neuronale pentru analiza parcărilor auto (Graf de rețea neuronală)

Abstract

În prezenta lucrare se va studia posibilitatea elaborării unui sistem, bazat pe rețelele neuronale în bază de graph, ce ar putea să prelucreze imaginele video, captate de pe camerele de luat vederi, și să determine cu o precizie înaltă locurile disponibile de parcare.

One of the most annoying jobs people have to do these days is park cars. Nowadays, finding a parking spot has become a problem that shouldn't be ignored because it takes time and effort. Finding a parking spot is a major hassle, especially in urban areas. This essay tries to develop one parking system that, in many ways, lessens parking problems. The method described in the study classifies parking spots in a park into empty and filled slots using a machine learning model called a convolution neural network (CNN). In this study, the Transfer Learning approach is used to optimize the categorization task. Parking issues touch more than just the drivers; they also have a significant negative impact on the environment and much broader and more widespread concerns. Therefore, it is crucial to have a parking system in place. The model suggested in the research drastically reduces the amount of time a motorist has wait for a vehicle by sending parking instructions far in advance.