LFLD-CLbased NET: A Curriculum-Learning-Based Deep Learning Network with Leap-Forward-Learning-Decay for Ship Detection

Ship detection in the maritime domain awareness field has seen a significant shift towards deep-learning-based techniques as mainstream approach. However, most existing ship models adopt random sampling strategy for training data, neglecting complexity differences among samples and learning progress of model, which hinders efficiency, robustness, generalization ability. To address this issue, we propose model called Leap-Forward-Learning-Decay Curriculum Learning-based Network (LFLD-CLbased NET). This incorporates innovative strategies curriculum to enhance its capabilities. The LFLD-CLbased NET is composed ResNet feature extraction unit, combined with difficulty generator scheduler. effectively expands data based on real ocean scenarios, scheduler constructs corresponding enabling be trained an orderly manner from easy difficult. strategy, allows flexible adjustment rate during training, proposed enhancing efficiency. Our experimental findings demonstrate that our achieved accuracy 86.635%, approximately 10% higher than other models. In addition, conducted extensive supplementary experiments evaluate effectiveness tasks. Furthermore, exploratory different modules compare performance under varying parameter configurations.