Optimizing Rock and Mine Detection using Deep Learning and Advanced Classification Models

Rock and mine detection is crucial for resource extraction and other practices. The existing technologies are not efficient because of their complex structure and are prone to noise quality issues. The proposed technique comprises of Deep Learning (DL) model and an advanced Multi-Attention Residual Network (MARN) for optimal rock and mine detection. The impurities present in the geological data obtained from the SONAR dataset are removed by data grooming, and the input data are transformed based on the requirements of DL model, which is done by data shaping. The data exploration identifies potential key patterns, anomalies, and trends. The data is split into a training and a testing model. The optimal features used to enhance the system performance are selected by feature selection. The proposed MARN accurately detects the rock and mine types, and it is implemented in Python software. Our proposed model achieves 92.85% accuracy, 100% precision, 88% recall, and 94% F1-score based on the experimental results.