Modeling the number of people in Venice during COVID-19: a State Space approach

This thesis investigates the effectiveness of protective measures during COVID-19. Specifically, this work is about the use of Dynamic Linear models (DLMs) to analyze the number of people present in photos taken at Campo San Felice in Venice. To obtain this time series, Yolov7, a state of the art model in object detection, was trained on a custom dataset and used to extract the number and the position of people from each photo. Additionally, with this data, the DBSCAN algorithm was applied to identify and analyze the clusters of people present in the different pictures. The DLMs were then fitted to the data and compared to Prophet, a modular regression model developed by Meta. The study found that the DLM was effective in capturing the trends and seasonality of the time series, and highlighted the potential of using DLMs in time series analysis, particularly for analyzing real and complex data.