@article{GUO2025, 
author = {Bai'en GUO and Fushi CHEN and Zhifei ZHOU and Yao SHEN and Yiyi LI},
title = {An Automatic Annotation Algorithm for Shooting Bullet Trace Images Based on High-resolution Network},
year = {2025},
journal = {Forensic Science and Technology},
volume = {50},
number = {3},
pages = {252-258},
keywords = {gunshot feature extraction, deep learning, high-resolution networks, image recognition, automatic labeling},
url = {https://www.sciopen.com/article/10.16467/j.1008-3650.2024.0042},
doi = {10.16467/j.1008-3650.2024.0042},
abstract = {The identification of the characteristics of shooting bullet trace image is the main content of gunshot trace inspection, and also one of the challenges. This article introduces an advanced automatic annotation method for shooting bullet trace features based on the High-resolution networks(HRNet) framework, which can achieve automatic labeling of the land-engraved trace area, groove-engraved trace area, and slippage trace area. A database of 5 985 images containing seven different sizes of shooting bullet traces extracted by BalScan(3D trace image scanning system) was constructed and divided into training, validation, and testing datasets at a ratio of 7:1.5:1.5. The training dataset was manually annotated to identify the land-engraved trace area, groove-engraved trace area, and slippage trace area, which were used to train the high-resolution network model. Then, the unlabeled testing dataset was input into the trained model for automatic annotation of the feature areas. Finally, the annotation results were manually reviewed and the accuracy was recorded. The results showed that the proposed method achieved an average accuracy of 94.1% in the automatic annotation task, demonstrating its effectiveness. This annotation algorithm for shooting bullet trace images without manual annotation can significantly reduce the workload of inspectors and provide a feasible new approach to improve the efficiency of firearm trace inspection.}
}