AI-Track-tive: open-source software for automated recognition and counting of surface semi-tracks using computer vision (artificial intelligence)
Simon Nachtergaele, Johan De Grave
Abstract
Abstract. A new method for automatic counting of etched fission tracks in minerals is described and presented in this article. Artificial intelligence techniques such as deep neural networks and computer vision were trained to detect fission surface semi-tracks on images. The deep neural networks can be used in an open-source computer program for semi-automated fission track dating called “AI-Track-tive”. Our custom-trained deep neural networks use the YOLOv3 object detection algorithm, which is currently one of the most powerful and fastest object recognition algorithms. The developed program successfully finds most of the fission tracks in the microscope images; however, the user still needs to supervise the automatic counting. The presented deep neural networks have high precision for apatite (97 %) and mica (98 %). Recall values are lower for apatite (86 %) than for mica (91 %). The application can be used online at https://ai-track-tive.ugent.be (last access: 29 June 2021), or it can be downloaded as an offline application for Windows.