Oral Presentation IPWEA International Public Works Conference 2025

The Victorian Department of Transport and Planning (DTP) is harnessing advanced artificial intelligence (AI), deep learning, and geospatial technologies to address critical data gaps in its road asset management framework. In collaboration with Fedasen, DTP implemented Fedasen’s intelligent Road Analysor (FiRA), which integrates high-resolution aerial imagery and 360-degree ground-based video feeds to automate the detection, classification, and condition assessment of road assets that previously lacked comprehensive baseline information. FiRA’s object detection models, trained and fine-tuned on large, diverse Fedasen datasets, enable robust asset identification and severity analysis through video processing. These outputs, in conjunction with other sensors and geospatial data, feed georeferenced asset insights into a Geographic Information System (GIS), providing detailed locational data for each asset subtype. The two most important aspects of data quality that have been addressed through the implementation of the AI-powered approach are the accuracy and precision of the data, as well as the completeness and coverage of assets. Using this approach, within 18 months, nearly 2.5 million assets encompassing 138 asset types and subtypes across six asset classes defined in DTP’s asset hierarchy standard were classified and registered. A multilayer, multiparty error detection and data quality validation approach with humans in the loop and AI as an assistant for data validation has been implemented, demonstrating an overall accuracy of 99.96%. This paper presents a case study on the implementation of FiRA AI technology for DTP, comparing it with other approaches and evaluating its effectiveness in meeting the asset hierarchy standards and asset data quality requirements targeted by DTP.