Changes between Version 17 and Version 18 of Other/Summer/2025/CityOS

Timestamp:

Jul 17, 2025, 4:57:54 AM (6 days ago)

Author:

sh1746

Comment:

—

Other/Summer/2025/CityOS

@@ -4 +4 @@
 === Project Information
 
-''City''OS is a project organized and developed under the guidance of Dr. Jorge Ortiz at WINLAB, Rutgers University. The entire goal for this project is to research how privacy is handled within data collection as groundwork for future privacy infrastructure within the contexts of smart cities and urban living. Currently, the methods of data collection are separated into three main API's, which the teams under Dr. Ortiz have been working on before our arrival. The first API, also called the single locality data aggregation, is where the data is collected, used, and then discarded in a continuous fashion
+''City''OS is a project organized and developed under the guidance of Dr. Jorge Ortiz at WINLAB, Rutgers University. The entire goal for this project is to research how privacy is handled within data collection as groundwork for future privacy infrastructure within the contexts of smart cities and urban living. Currently, the methods of data collection are separated into three main API's, which the teams under Dr. Ortiz have been working on before our arrival. The first API, also known as On-Scene API, is where data is collected, used, and then discarded in a real-time, continuous manner. The second API, called the Single-Locality Aggregation, is a continuous data collection that is fixed to a small (localized) region. The third and final API, called the Cross-Locality Aggregation, is a citywide collection incorporating data from multiple localities and user devices.
 
+For our summer internship, we will be focusing on the second API - The Single-Locality Aggregation. 
 
-Our key mission is to leverage computer vision and machine learning algorithms to create a smart parking system that predicts parking availability, reduces congestion, and improves urban quality of life.
+To demonstrate how the API functions, we will be creating an end-to-end parking application that will continuously aggregate information of the parking spots here at WINLAB to train a machine learning model to predict future parking availability. 
+
+We hope this research and application will lay the groundwork for a larger-scoped data-driven citywide parking app, which will reduce vehicle congestion and improve urban quality of life. 
 
 === Documentation
-  Our parking predictor consists of 5 main components, which are the following: the Data Collection, Video Processing, Data Formatting, Prediction Modeling, and the User Interface.
+  Our parking predictor application consists of 5 main components, which are the following: the Data Collection, Video Processing, Data Formatting, Prediction Modeling, and the User Interface.
 
-|| [[Image(CityOS-Design Flowchart.png, height=500vw)]] ||  The data collection is being done from the Rutger's Winlab Camera which is pointed into the adjacent parking lot. It collects video footage daily from 8 am to 8 pm.  The video processing is where the video footage is analyzed for information. In our case, it is checking each of our 10 parking spots for availability. When a vehicle parks in any of the spots, it will return data that this spot has been occupied. We accomplished this functionality using OpenCV and YOLOv8.  The video formatting is where the video data is formatted and aggregated into a CSV file. Within a line, it displays the spotID number, Month, Day, Year, Hour, Minute, Second, and the status of whether the spot is occupied or free.  The prediction model is where the aggregated data is imported in order to train and teach a machine learning model. In our project, we are using a Random Forest model, which will take the data and use it to predict the occupancy given a spotID and a time.  The user interface is where people can interact with our systems. We are planning to implement a large interactive UI where users are able to see a mock replica of the parking lot and are able to then select and visualize parking lot occupancy over the days and weeks. ||
+|| [[Image(CityOS-Design Flowchart.png, height=500vw)]] ||  The data collection is being done from the Rutger's Winlab Camera which is pointed into the adjacent parking lot. It collects video footage daily from 7 am to 7 pm.  The video processing is where the video footage is analyzed for information. In our case, it is checking each of our 10 parking spots for availability. When a vehicle parks in any of the spots, it will return data that this spot has been occupied. We accomplished this functionality using OpenCV and YOLOv8.  The video formatting is where the video data is formatted and aggregated into a CSV file. Within a line, it displays the spotID number, Month, Day, Year, Hour, Minute, Second, and the status of whether the spot is occupied or free.  The prediction model is where the aggregated data is imported in order to train and teach a machine learning model. In our project, we are using a Random Forest model, which will take the data and use it to predict the occupancy given a spotID and a time.  The user interface is where people can interact with our systems. We are planning to implement a large interactive UI where users are able to see a mock replica of the parking lot and are able to then select and visualize parking lot occupancy over the days and weeks. ||
+||  ||  ||
 
 === Weekly Progress