Estimating the cost based on the damage for a given vehicle for insurance processing is a significant area and having huge scope for automation. At the point when a car gets harmed in a mishap, the insurer of the corresponding vehicle needs to take care of the expense. Human intervention in this process is costly and takes more time for visual inspection. By utilizing the advanced deep learning procedures, it is quite easy to prepare a model that will recognize the damaged car parts and estimate the cost accordingly. Convolutional neural network (CNN) is one kind of artificial intelligence technique commonly used for pattern recognition and classification. To get an accurate result, the model should be capable of anticipating what sort of maintenance precisely to follow. The model will be prepared by training the variety of images collected with different viewpoints of the same damaged vehicle before making the final predictions. We used transfer learning-based pre-trained VGG16 network architecture for this purpose. Though these algorithms are giving accurate predictions, memetic algorithms can still optimize the results. In addition to this, two other applications for detecting the car existence in a given image using deep learning useful for automated self-driving applications are presented. © 2020, Springer Nature Singapore Pte Ltd.