AI and deep learning for identifying pavement failures in Latin American and the Caribbean


Image courtesy Michael Busch on Unsplash

Commercial approach(es) used to catalyse investment: Adoption of an innovative digital tool based on artificial intelligence (AI) for the roads network assessment and planning processes; competitive advantage of reducing investment and time required by 98%.  

Technology approach(es) used to catalyse investment: An innovative tool that improves paved road network planning and reduces maintenance costs for government agencies, replacing a costly and lengthy manual process; AI application, digitalisation, and simplification of road assessment process

Finance approach(es) used to catalyse investment: Technical cooperation funds and loans to governments in Latin America and the Caribbean to deploy technological infrastructure (cloud-based, or on-premises if required), and to adapt, operate, maintain, and upgrade the model, as well as keeping it cyber-resilient. 

Key benefits: 

  • Capex efficiency 
  • Opex efficiency 
  • Climate mitigation 

Additional benefits

Improved road network maintenance planning to minimise interventions and reduce their environmental impact 

Scale of deployment: 

There are 627,418 kilometres of paved roads in Latin America and the Caribbean (LAC), serving a population of more than 565 million people  

Project value: 

Based on first assessments, an estimated investment of USD60,000 for a network of about 10,000 kilometres. This may extrapolate to USD3.6 million for the entire paved road network in the LAC region. 

Project start/end dates: 

Q4 2022 to Q4 2024 

Current status of the project: 

Operational. The app has been published in the IDB library and is available for deployment and improvement. 

Pavimenta2 is a platform developed by the Inter-American Development Bank (IDB) to identify, measure, and quantify pavement failures and to validate transportation signage. By simply driving through the roadway network with a mounted cell phone or GoPro®, Pavimenta2 uses computer vision technology combined with artificial intelligence (AI) and deep learning to automate the measurement of quantities and locations of blurred lines, linear cracking, transversal cracking, crocodile cracking, rutting, and other failures and uploads the recorded video.  The tool also identifies roadway signage, classifies each sign, and decides if it is in good condition or may need maintenance. 

Pavimenta2 automates the processes of documenting, measuring, and recording failures, speeding up data collection and delivering cost savings over traditional approaches – thereby alleviating qualified professionals, who are often in high demand, to focus on strategy and planning. The app is published in IDB’s library and already trained. According to the licence, anyone can do additional training for defect and signage detection.

Currently, assessing 10,000 km of a road network is a labour-intensive process that requires 18 months and costs an estimated USD3.2 million. With Pavimenta2, this assessment takes two weeks for an estimated cost of USD60,000 and enables IDB to identify transport and pavement patterns to create sector efficiencies.


IDB has been running pilots in Argentina, Brazil, Uruguay and five other countries in Central America.  

Challenges experienced/overcome in implementation

  • Limited access to funds: To finance the initial stages of product development from concept, MVP, model training, documentation, quality tests up until final publication in the bank library, IDB deployed its own staff and used technical cooperation funds and loans for hiring part time developers including for the initial roll-out.
  • Scaling up the tool: Requires developing a partnership with each of the numerous transport agencies at both national and sub-national level. 
  • Cloud infrastructure: The client's cloud infrastructure could be different from the developer's infrastructure. Moving the solution between different cloud platforms can be tricky and sometimes implies a whole new development. Support from local technical staff that know the client’s infrastructure is a good practice. Adaptation may include the adequation of client’s IT infrastructure, development of front-end (visualisations and reports) and back-end (cloud architecture, integration with existing systems) modules, training and documentation, maintenance, and cybersecurity.

Other approaches that enabled investment

Synergies with road infrastructure programs where the loan includes a component for institutional development. This tool may help to prioritise maintenance expenditure and future major capital interventions within the road network, while gaining efficiencies and minimising waste and emissions from construction activities.


Note: This case study and all information within was submitted by the Inter-American Development Bank Group (IDB) in response to our global call for InfraTech case studies. 

Last Updated: 21 October 2022