Circuit Help’s journey to creating an algorithm for vehicle counting and classification started when the entire country was placed in a lockdown in March 2020. Since physical distancing was implemented, traffic surveys involving manual count are practically impossible. Our experience in computer vision gave us the starting post to pursue the challenges to developed an AI that will actually work in the Philippines settings.

Using CCTV and cellphone footages, we started collecting our training data sets and built our first model to power our proof of concept. We then repeatedly retrained our AI as we used it in actual surveys making it more accurate. Our latest model was trained using over 250,000 individual training images and has around a 95% counting accuracy. This algorithm will further improve as usage is increased and data sets are expanded.

Since our Technology Demonstration with DPWH in Leyte last February 2021, our locally developed Artificial Intelligence (AI) have counted and classified over a million vehicles during traffic surveys in the following areas:

Leyte

Cavite

Quezon

Bataan

Zambales

SBMA

Pampanga

​