2  Summary

Assessing road quality provides valuable insights into the state of a country’s infrastructure and transportation system. By linking this data to existing datasets on road maintenance expenditure and traffic volumes, the outputs of this project can inform decision making and improve the efficiency of road maintenance and transportation planning. The assessment of road quality is crucial for economic development and efficient transportation in South Africa.

2.1 Purpose

Conduct a first stage quantification of the quality of roads in South Africa using Google Streetview imagery. The computer vision pipeline could be exploited with real-time images captured by municipal vehicles to improve monitoring and maintenance.

2.2 Timeline:

  • Project scoping discussions with stakeholders: 1 month
  • Data collection and annotation: 2 months
  • Model development and training: 2 months
  • Model testing and evaluation: 1 month
  • Implementation and analysis: 1 month
  • Report writing: 1 month
  • Handover of productionized computer vision model: 1 month

2.3 Data Requirements:

  • Google Streetview images of South African roads
  • Labelled and annotated images of various road features such as potholes, cracks, and debris

2.4 Sampling

  • Random sampling across one region of the country, or
  • Stratified sampling to evaluate the quality of roads in different municipalities.

2.5 Computing Requirements:

  • High-performance computing resources such as a GPU-enabled machine for model training and inference (Google colab +)
  • A large storage capacity to store the images and annotation data (Big Google drive will be fine)

2.6 Output:

  • A computer vision model that can automatically identify and quantify various road features such as potholes, cracks, and debris.
  • A report that includes the findings from the analysis of the model’s output, including the overall road quality, the distribution of road features, and the impact on traffic and safety.
  • The output would be useful for quantifying the quality of South Africa’s roads, which is a valuable asset for the nation. The model’s output can be used to identify areas in need of repairs and maintenance, prioritize resources, and track progress over time. Additionally, the report can be used to inform policy decisions and allocate funds for road maintenance and improvement.