Geostore

Over 13,000 road projects with 200,000+ images from four different government agencies are tracked in the Geostore

Project

A Disclosure, Monitoring, and Feedback platform for tracking public road projects

Services

Web Application Development, Product Study, Strategy Design

Visit Website

Over the years, the Philippines has been significantly increasing the level of public investment for local roads. This dramatic increase introduced a need to transparently track these infrastructure projects and provide visibility to the government agencies, development partners and to the Filipino people.

To provide a central repository for all the data collected in road projects, with the support of the World Bank, the Philippine government partnered with Symph to develop the Geostore, an open source platform designed to be the central database of all geotagged information for roads within the Philippine archipelago.

 

The Challenge

Recently, government agencies began tracking road projects using geotagging technology. Geotagging refers to the systematic collection of images and/or videos along location referenced tracks. However, the geotagged data was just managed in the individual agencies and there was little to no collaboration of the geotagged data between agencies. Likewise, with the expansion of road projects within the nation, the constant generation of geotagged data across agencies was becoming cumbersome to manage and keep organized.

These challenges sprouted the concept of the Geostore, a platform that organizes and displays information on the government roadwork projects in a useful manner.

Responsive image

Our Approach

Through a series of collaborative meetings and workshops with the concerned government offices and the World Bank, we were able to get a more accurate grasp of how the system currently works, and how it can be improved for more efficiency and for a more effective collaboration between the agencies involved in the process.

Furthermore, Symph recognized an opportunity to improve the road classification process that was used to be handled manually by platform users. We developed a machine learning model that could classify the road surface types faster. We used approximately 10,000 images that were manually identified by humans to train the machine learning model, which then processed the full set of over 200,000 images with an 84% accuracy in a fraction of the time it would take human users.

Responsive image

Our Output

Geostore Dashboard and Viewer​

The Geostore dashboard is where government agencies can easily upload the geotagged data from road projects – project details, images, and other important information. The dashboard also provides them with the ability to easily track and monitor road projects.

The public can easily view all of the uploaded projects in the Geostore viewer.

Road Type and Quality Classifier

Once uploaded to the dashboard, the images and tracks are geoprocessed, a term we coined which means converting this information into tabular and spatial data concerning the route, surface type, surface quality, and width of road segments. This is an important step to keep accurate track of the state of our public roads.

Because of the the machine learning model we developed, the Geostore platform can also classify road types with greater efficiency.

Responsive image

More work

Next Page Our Work