Bullet impact images dataset on different materials for forensic ballistics applications

Description of the project. This project aims to advance the field of forensic ballistics by creating a large, controlled dataset of bullet impact images to train deep learning models for automatic trajectory reconstruction. The dataset was generated in collaboration with the Criminalistics Service of the Spanish Civil Guard to ensure real-world relevance and accuracy. Without this collaboration, this dataset would not have been possible. Description of the dataset. The dataset consists of 1,800 high-resolution images (in PNG format) of bullet impact holes, captured under controlled laboratory conditions. It covers six drift angles (15°, 30°, 45°, 60°, 75°, 90°) on two common building materials, frequently encountered in forensic investigations: particleboard and drywall. Each angle category contains 300 images, equally split between the two materials (150 images for particleboard and 150 images for drywall).