In forensic anthropology, skeletal structures play an important role in sex identification. The cranium and pelvis bones provide a higher accuracy rate in identification in terms of sexual dimorphism. The calvaria, which is a part of the cranium, is an important structure in terms of sex identification because it shows sex-related differences in shape, structure, and size. This study aims to estimate sex with deep learning (DL) models based on calvaria photographs. In the study, calvaria photographs of autopsied cases over the age of 18 were analyzed. We analyzed 210 photographs of the inner (endocranial) surface (105 male, 105 female) and 310 photographs of the outer (ectocranial) surface (155 male, 155 female). Calvaria photographs were trained and tested with Support Vector Machine (SVM) and K-Nearest Neighbor (K-NN) algorithms using attributes obtained with the ResNet50 DL model. As a result, the highest success rate for sex estimation from endocranial photographs was 96.43% with Fine K-NN, while the most successful model for estimation from ectocranial photographs was Cubic SVM with 96.77% accuracy. Successful results were obtained in the sex estimation study performed with DL models directly from calvaria photographs without morphometric measurements. Future studies are needed to improve the performance of DL models.