In Turkey, lots of car accidents occur every day. Due to the number of accidents happening, expertises must put significant effort into the damage estimation process. Such a process includes many manual tasks and paperwork that takes a long time to complete and might not be error-free. Luckily, the latest development on computer vision systems makes it applicable to automate such processes. An automatic claim recognition system takes a significant role in insurance systems to reduce the required amount of time and human-based errors. In this paper, we focused on detecting and segmenting the car parts from car images which is a necessary component of claim recognition systems. To segment car parts correctly, a system with two stages has been designed. In the first stage, the angle of the car is classified. This information is later used in the system to decide which car part segmentation model to use. In the second stage, the same information is pieced together with the car part segmentation model output to segment the exact car part. For car part segmentation, a deep learning model, Unet, has been used. The results of the system satisfy the determined metrics and prove that the system is applicable to automatic claim recognitions.