Pothole Detection using TINYML

Pothole Detection using TINYML

Sanathkumar S J1, Ramalingam H M2, Yajnesh K3, Sheshank Kulkarni4

Student, Electronics & Communication Engineering, MITE, Moodabidri, India1

Senior Assistant Professor, Electronics & Communication Engineering, MITE, Moodabidri, India2

Student, Electronics & Communication Engineering, MITE, Moodabidri, India3

Student, Electronics & Communication Engineering, MITE, Moodabidri, India4

---------------------------------------------------------------------***---------------------------------------------------------------------

Abstract - Potholes present a widespread issue on roadways globally, causing vehicle damage, accidents, and posing safety concerns for drivers and pedestrians. Detecting these potholes involves identifying and pinpointing their locations, a process traditionally performed manually by road inspectors, which is both time-consuming and ineffective. The complexity of pothole detection arises from their varying sizes, shapes, and depths, often obscured by shadows, debris, or other road features, making them challenging to discern. However, automated pothole detection systems are gaining popularity due to their ability to swiftly and accurately identify potholes across extensive areas. As technology advances, particularly with the refinement of machine learning algorithms, these automated systems are becoming increasingly precise and efficient. This progress holds promise for enhancing road safety and reducing the financial burden associated with road maintenance.

Key Words:  TinyML, Edge-computing, Fomo mobileNet v2, Pothole detection.