Oracle bone inscriptions constitute a core origin of Chinese civilization. This paper presents OBWX, a dataset that collects and organizes glyph images of Chinese ideographs across four critical historical periods: oracle bone inscriptions, bronze inscriptions, warring states inscriptions, and small seal inscriptions. The dataset comprises 74,047 images covering 6,650 ideograph categories. Through benchmark experiments on multi-period character classification, we validate the usability of multiple mainstream deep learning models on this dataset. The DenseNet-121 model achieves a classification accuracy of 91.58\%, demonstrating the practical utility of OBWX for ancient ideograph recognition. The OBWX dataset provides a standardized, high-quality data foundation for intelligent research on ancient ideographs and holds promise for fostering further breakthroughs in digital philology.