Page 82 - AC-2-2

Arts & Communication Digital reconstitution of lost heritage

selected paintings. Initial steps involved embossing the and 0.05, respectively.
original images by adjusting brightness and contrast
settings, preserving brush strokes’ visibility while 5. Recommendations
mitigating background noise. These embossed renditions Drawing on the results from the digital recreation of the
served as bump maps and were processed through a 1785 Salon, focusing on Jean-Joseph Taillasson’s Philoctete
dedicated normal map generator (https://cpetry.github. à qui Ulisse & Néoptolème enlèvent les Alèches d’Hercule and
io/NormalMap-Online/; accessed January 6, 2024). the AI reconstruction of Antoine Vestier’s La nonchalante,
Specific settings included: a strength value of 2; a level several key recommendations emerge to guide future
setting of 8; blur/sharpness intensity set to 2; utilization of endeavors in the digital reconstruction of cultural heritage.
the Sobel filter; and a displacement parameter set at 0.2. In the reproduction of existing artworks like Taillasson’s,
The generated normal maps, accompanied by ambient the employment of advanced image editing techniques is
occlusion and specular information, were seamlessly paramount. Addressing issues such as glare, stretch, and
integrated into the material setup within the Unreal Engine skew in the original photographs ensures a more accurate
framework. Further, adjustments ensued, flattening the foundation for subsequent digital processes. The creation
map by 5% to refine its visual impact. Additional material and utilization of bump and normal maps from these
properties were configured, setting metallic values to 0.3, edited images are critical steps. These techniques allow for
the successful replication of the textural qualities of the
while both specular and roughness were fine-tuned to 0
original paintings, contributing significantly to the realism
and authenticity of the digital replicas.
Furthermore, the project’s success in utilizing AI for
the reconstruction of Vestier’s lost work underscores
the potential of AI in reconstituting missing cultural
artifacts. This approach can be expanded by incorporating
photogrammetry to recreate lost materials. The
combination of AI and photogrammetry, underpinned by
rigorous art historical research and methodologies, opens
new avenues for reconstructing materials that have not
survived to the present day. This method not only aids in
the visual reconstruction of artworks but also contributes
to our understanding of the materiality and techniques
used in the original creations. These strategies, blending
technological innovation with traditional art historical
Figure 3. Pietro Antonio Martini, View of the Salon of 1785, engraving.
The Metropolitan Museum of Art, New York. Creative Commons Zero, research, represent a substantial advancement in the field of
Public Domain Dedication. digital humanities. They not only facilitate the preservation

Figure 4. Jean-Joseph Taillasson, Philoctete à qui Ulisse & Néoptolème enlèvent les Alèches d’Hercule, 1784. (Left) Picture edited to remove glare, stretch,
and skew. (Center) Bump map from the original picture. (Right) Normal map generated from the bump map. Imagery created with Unreal Engine 5 and
Quixel Megascans.

Volume 2 Issue 2 (2024) 8 doi: 10.36922/ac.2719