Machine Learning and Computer Vision Consulting Services

The vendor is required to provide machine learning and computer vision consulting services for this project will include the following:
•    Video log data and lidar acquisition, 
•    Data labeling, 
•    Model training and validation, 
•    Extraction of roadside features
•    Visualization and integration with GIS, 
•    Technical report.
1. Data acquisition
•    Video log imagery
•    GIS shapefile of the agency’s current linear referencing system
•    GIS shapefile and table that identifies the roads for collection, along with all collocated nondominant routes for omission
2. Data labeling
•    The service provider will manually annotate a sample set of roadway images covering rural and state highways, from a diverse set of rural conditions in state, by labeling the roadside features listed previously. 
•    The service provider will perform image preprocessing (resizing, color correction, noise removal) as necessary.
3. Model training and validation
•    The service provider will train machine learning models for each roadside feature data element using the annotated set of imagery. 
•    The models should be able to geo-localize the roadside features. 
•    The service provider will split the annotated set into 80% for training and 20% for validation. 
•    The models are to be validated using various performance measures including accuracy, performance and recall, and other methods of evaluating performance where appropriate. 
•    The performance should be at least 75% accuracy in detecting, classifying or measuring the roadside feature data element.
4. Extraction of roadside features
•    The service provider will apply the trained models to the rural federal-aid highway network and detect, classify, and geo-localize the roadside feature data elements.
5. Visualization and integration with GIS
•    The service provider will integrate the results of the data extraction for each element on the linear reference system in an ArcGIS map. 
•    Applicants should propose additional ways of visualizing and interacting with the results.
6. Final technical report
•    The report will follow this general outline: abstract, introduction, data acquisition, data processing, data labeling, model training, validation and performance, statewide implementation, visualization and GIS integration, conclusions, opportunities for improvements.

- Contract Period/Term: 1 year
- Questions/Inquires Deadline: October 10, 2025