Enhancing Character Detection: Data Augmentation, Secondary Labeling, and Oracle Bone Script Filtering

To enhance the accuracy of our character detection model, we implemented several key improvements. First, we manually expanded the dataset, incorporating additional training data specifically for challenging characters. This targeted approach aimed to strengthen the model's ability to identify difficult-to-detect characters. Second, we conducted secondary labeling of the dataset to enhance its reliability and credibility. Finally, to address the issue of misclassifying characters similar to oracle bone script, we introduced an oracle bone script recognition-assisted detection algorithm. This algorithm effectively filters out incorrect prediction results, improving the overall accuracy of the character detection process.