One of the most common types of road damages are the potholes and cracks leading to increasing risk of the road accidents and cost of maintenance. The manual surveys are time consuming and can cost a lot of money and are also not up to the mark. To bypass all the problem the object detection models like mobilenet or yolo can be finetuned for the road damage detection using custom dataset and deploying them onto the mobile devices enabling the on device processing bypassing the server side processing that is expensive. Methodology: ● Choose any relevant object detection model. ● Create a custom dataset to train the model for as many types of classes you want. ● Train the model for mobile devices in nano version and export this in the TFlite or ONNX format. ● Check for the output and input of the model which provides us with the shape of data for both input and output tensors. ● The model can be trained on two types of images: grayscale and RGB. The model can be used in the mobile app by using the tensorflow.js format of the model which gives us the model in the json format and its weight in a bin file and a .yaml file . These files are necessary to create a fully working platform. Now select the framework which we need to develop for the app here we choose the react native or flutter.