Philippine License Plate Detection on Mobile

License plate detection is an essential part of Automatic License Plate Recognition (ALPR) systems. While there are already such systems in other countries, the use of ALPR is still not as prevalent in the local Philippine setting. There have been previous plate detection systems on Philippine datasets, but since then rapid advances in machine learning have offered new alternatives for object detection models that have the potential for performing at greater efficiencies. This study aims to evaluate different object detection models for the development of license plate detection in the Philippines, particularly for mobile use. These models are trained on a custom Philippine license plate dataset. A more optimized and lighter solution is more convenient to use and deploy, allowing its potential utilization in a mobile setting. The models that were evaluated were tested for accuracy in terms of mAP, inference speed, and model size to determine their capability to perform efficiently for mobile use.