明成化斗彩鸡缸杯：釉下青花与釉上彩差值对齐AI鉴别数据集

1. “双子互证”式AI模型校验：利用本数据集独有的**“成对锚点”（Cup A 与 Cup B），训练 Siamese Network（孪生神经网络）。通过对比两只杯子在“斗彩对齐度” (Doucai_Alignment_Score)** 上的细微同源差异，建立成化御窑的“工手一致性”模型。这能有效识别出市面上“单只极精，但成对时风格不统一”的高端拼凑伪作。 2. 斗彩工艺的数字化拆解与复原：基于短波红外与可见光的分层数据，在数字空间中将“釉下青花”与“釉上五彩”进行图层剥离。结合**“青花下沉度” (Subglaze_Blue_Depth)** 数据，真实还原成化官窑“釉下青花为骨，釉上彩料为衣”的立体工艺结构，可用于开发高保真的 VR 鉴赏应用。 3. 微观特征的全生命周期监控：针对鸡缸杯极高的价值，利用数据集中的**“姹紫哑光度” (Purple_Glaze_Matte)** 和 “青花下沉度” 建立微观指纹档案。在文物借展归还或运输交割时，通过便携设备快速复扫这两个指标。一旦发生替换（高仿品的青花深度和紫彩光泽度必然不同），数据将产生剧烈偏差，从而实现定责理赔。

1. Dual Mutual Verification-based AI Model Validation: Leveraging the unique **Paired Anchors (Cup A and Cup B)** in this dataset, we train a Siamese Network. By comparing the subtle homologous differences between the two cups in terms of *Doucai_Alignment_Score*, we establish a craftsmanship consistency model for Chenghua Imperial Kilns. This can effectively identify high-end counterfeit pieces that are "excellent individually but stylistically inconsistent when paired" in the current market.
2. Digital Disassembly and Restoration of Doucai Craftsmanship: Based on layered data collected via short-wave infrared and visible light imaging, we separate the "underglaze blue" and "overglaze five-color" layers in the digital space. Combined with the *Subglaze_Blue_Depth* data, we faithfully restore the three-dimensional craft structure of Chenghua Imperial Kilns, which follows the principle of "underglaze blue as the framework, overglaze pigments as the coating". This can be used to develop high-fidelity VR appreciation applications.
3. Full-Life-Cycle Monitoring of Microscopic Features: Given the extremely high value of the Chenghua Chicken Cup, we establish a microscopic fingerprint archive using the *Purple_Glaze_Matte* and *Subglaze_Blue_Depth* metrics from the dataset. During the return of loaned cultural relics or transportation handover, these two indicators can be quickly rescanned via portable devices. If replacement occurs (the blue pigment depth and purple glaze luster of high-imitation products are inevitably different), the data will show drastic deviations, enabling responsibility determination and claims settlement.

搜集汇总

背景与挑战

背景概述

该数据集专注于明成化斗彩鸡缸杯的AI鉴别，通过提供成对锚点训练孪生神经网络，实现‘双子互证’式校验，有效识别高端拼凑伪作。同时，它利用分层数据数字化拆解釉下青花与釉上彩工艺，还原立体结构，并基于微观特征建立指纹档案，支持文物全生命周期监控和定责应用。