Project: #335 Real-Time Traffic Analytics at Intersections Progress Report - Reporting Period Ending: March 31, 2021 Principal Investigator: Srinivasa Narasimhan Status: Active Start Date: July 1, 2020 End Date: June 30, 2021 Research Type: Applied Grant Type: Research Grant Program: FAST Act - Mobility National (2016 - 2022) Grant Cycle: 2020 Mobility21 UTC Progress Report (Last Updated: March 30, 2021, 10:52 a.m.) % Project Completed to Date: 100 % Grant Award Expended: 100 % Match Expended & Document: 100 USDOT Requirements Accomplishments The major goals for this project are 1) develop methods to detect anomalous behavior in the street environment, 2) provide an easy-to-visualize web-based platform for viewing live video feeds and analytics, 3) deploy cameras at another intersection, 4) make data and results available to the public. Accomplishments under the above goals include: 1) methods were developed to perform 4D reconstruction with a single view camera and automatically detect anomalous vehicular activity, 2) a website has been built to show live feeds from the cameras, preliminary display of analytics are on the site, but currently switching to tableau, 3) data collection underway at the intersection of Fifth Ave and Morewood following a collision between an vehicle and a pedestrian last Winter, 4) paper submitted to a top-tier IEEE conference (IVS). Multiple students from the MSCV program have worked on this project giving them a hands-on, real-world experience to prepare for their research future in industry. Results are disseminated via publications and a project webpage. Impacts Deployment of platform and data collection has given rise to algorithm development that has led to high level analytics describing anomalous vehicular activity at intersections. Other Collection of large amounts of a unique data set including different weather conditions at different times of day. Data being used to train a a generalized DNN model for detecting vehicles and people. Publications New algorithms for 4D reconstruction from a single view camera and automatic anomaly detection Website (http://platformpgh.cs.cmu.edu) updated to provide project information, live streams, and analytics for deployment sites. Outcomes New Partners We have partnered with Radium who is providing cloud-based resources for training Deep Learning Models Issues No significant changes or problems