This project will develop a novel computational platform that will automate the entire process of reviewing, finding, extracting, and reporting engineering details from bridge plans. Work in Stage 1 focusses on automating the data and feature extraction process from drawings and tables from bridge plans using state-of-the-art deep learning algorithms. A holistic review of a variety of bridge plans will be performed to identify and categorize the engineering details of interest. This will be followed by detection of the physical objects of interest in the bridge plans, identifying the types of the objects, and extracting their main dimensions and details. This is to be achieved through the utilizing the capabilities of the CNN algorithms that automate the entire process after initial training on a set of bridge plans. The details of interest can vary from geometric dimensions (e.g., height and width) to reinforcement properties (e.g., size and spacing of longitudinal and transverse bars). Next, post-processing operations will be developed for the extracted raw data and features to report them in desired output formats. Further to drawings, bridge plans often contain various tables with important information about a variety of engineering details. These tables will be located, the boundaries of each table’s cells identified, necessary data points extracted, and reported in a desired editable format (such as an Excel spreadsheet). With the automated identification and transferring of data and features from bridge plan sets to spreadsheets, post-processing activities related to making queries or finding quantities of interest will be greatly facilitated. Work in Stage 2 will extend the sources of information used for data and feature extraction form drawings and tables (completed in Stage 1) to text blocks. Stage 2 work will also involve extensive testing, assessment, and quality control of the developed computational platform using a variety of bridge plan sets provided by the Iowa, Minnesota, and California DOTs. A process will be established for the extraction of textual information from bridge plans. The accuracy and speed of the algorithms developed for data and feature extraction from drawings, tables, and text blocks will be systematically assessed. This assessment will start from the verification stage to make sure that the automated platform returns correct outputs for the bridge plans used in the training of the algorithms and will span a variety of desired outputs with both single and multi-source characteristics. The quality control effort will then be extended to the validation stage, in which the developed platform will be tested on several bridge plans not used for training purposes. The generated outputs will be compared with those obtained from manual extraction to identify and properly address possible errors and bugs . The final report will provide all relevant data, methods, models, and conclusions along with guidance on how to use the developed computational platform to automatically extract the data and features of interest from bridge plan sets.