This article presents an approach applying computational intelligence for the problem of detecting and estimating car damage from images. An automatic approach is developed applying specific variants of convolutional neural networks, a domain decomposition approach, and exhaustive data preprocessing techniques. A specific real case study is addressed, using real information from the State Insurance Bank (Banco de Seguros del Estado, BSE) in Uruguay. Results demonstrate the effectiveness of the proposed approach for identifying and estimating damages in several vehicle components and the potential of applying machine learning techniques for automatic car damage estimation task for the considered case study.