A DEEP LEARNING-BASED REAL-TIME AUTOMATIC ROCK LOGGING MODEL
Rock logging is a vital method for obtaining geological and geotechnical information. However, the conventional approach of manual core logging is a labor-intensive and time-consuming task, relying heavily on expertise and domain knowledge. To address these challenges, we propose a deep learning-based object detection model for automated rock core logging. Based on the YOLO (You Only Look Once) structure, we incorporate the attention module into the backbone and finally establish the enhanced Att-YOLO model. The performance of this proposed model is evaluated using our customized rock dataset, which is compiled from Hong Kong ground investigation reports. The results show that the model achieves the highest accuracy among all tested algorithms while maintaining real-time detection speed.
Additional information: Mr. Sihao YU, sihaoyu@connect.hku.hk