Counting the Number of Potholes in a Video Sequence

Pothole detection holds significant importance for road maintenance and safety. However, accurately counting potholes within dynamic video sequences poses substantial challenges due to varying lighting conditions, camera perspectives, object occlusions and limited availability of training and testing of dataset. To address these challenges, a methodology based on Cosine similarity, leveraging visual pattern recognition to quantify the number of potholes. Our approach achieves a promising standard error count. This study contributes to enhancing the accuracy and efficiency of pothole detection and counting systems, crucial for ensuring road infrastructure integrity and reducing hazards like pothole deaths, traffic congestion, gross wastage of fuel.