Art Forensics – The thrill towards knowing Art.
The intersection of Art and Science has long been a subject of interest. It is evident across various fields of numeracy that captivate the learner’s attention. From the realms of interior design and architectural landscapes to the intricate design of Ferrari’s V6 engine, these disciplines are intertwined with both artistic and scientific principles to fulfill their primary objectives: promoting healthier living, creating visually striking skylines, and enhancing speed.
The Art Forensics project did not emerge as an instinctual concept, nor was it a spontaneous initiative to analyze images. In early 2016, our founder, Yuhui Lin, was afforded several opportunities to collaborate with the media on furniture designs in Singapore, Europe, and Japan. As mobile applications surged in popularity, The Waterhouse SG contemplated the integration of virtual reality (VR) with designer furnishings. Given the substantial size and weight of many items, spatial awareness became a critical factor. Although VR and Augmented Reality (AR) appeared to be a promising venture, subsequent data and demographic analyses indicated that they would not be a viable long-term industrial solution.
In 2023 and 2024, Keras360 made significant strides in the data science domain. Engaging in hands-on coding for the development of user interface dashboards and calculators for patient survival records, data science for Art Authentication emerged as a compelling intellectual pursuit. The project was formally established in September 2024, with an analytical framework developed by October 2024. The initial success of the Art Forensics (AF) tool was its capability to identify artworks from specific artists, such as the previously unseen masterpieces of Oscar Claude Monet, including his series on waterlilies (Provoke 3). Like many great masters, Monet produced several variations of identical subjects, each differing in hue, perspective, brushstroke technique, and seasonal context. It is unlikely that either Monet or Van Gogh anticipated that their habits of artistic refinement would ultimately serve as an advantage for our analysis.
The abbreviation “AF beta” refers to Art Forensics in its beta version, also known as an experimental project. As is typical with any enterprise initiative, risks were associated with the launch of this new endeavor. Initially, Y Lin evaluated several existing analytical software packages; however, the technical outputs failed to overcome existing limitations. A significant constraint arose from the differing behaviors of analyses in relation to patient image scans, such as pulmonary CT and X-ray scans, which bore little resemblance to capturing imagery of schools of fish in the depths of the ocean. Accuracy, categorization, and sensitivity are critical components that must focus on authentication.
In contrast to most enterprise initiatives, we commenced this project on a modest scale, catering to select clients who may necessitate authentication of artworks, artifacts, or specimens to foster confidence during sales presentations. Should you wish for Keras360 The Waterhouse to incorporate our tool and/or AF library into your sales and marketing efforts, please do not hesitate to contact us at contact.general@keras360.io.
visit : https://keras360.io/afbeta for the beta experiment videos.
Were there any particular artworks or artists that might have contributed to analytical design?
Certainly. The painting Salvator Mundi – Savior of the World represents a significant artwork that sparked the desire to unravel the mystery surrounding its true creator. It is worth noting that many esteemed artists have had exceptional pupils who may have contributed to various aspects of the masterpieces housed in private collections or presented at auction houses. Artworks such as the three versions of the Mona Lisa featured in the Keras360 Provoke 9 video clip, as well as the imitations of artistic styles by various artists during the 19th and 20th centuries (Provoke 6), and the mimicry seen in generative AI art (Provoke 2), served as reference materials utilized in the design and calibration of the AF beta tool.
About mimicry and Generative AI art
At Keras360 The Waterhouse, we actively pursue captivating and intellectually stimulating art images that elucidate the objectives of art forensics. Should your generative AI artwork have been exhibited, it is understood that your work is appropriate for machine learning applications and sufficiently robust to surpass most machine evaluations. In technical terms, a calibration is required for minor adjustments. In layman’s terms, this process entails a refinement that allows the tool to establish an optimal position for the parameters, thereby yielding representative reference values such as ‘Found match’ and ‘Image not found,’ among others.
The Million-Dollar Question: What Are the Limitations of the Tool, and Are There Technical Solutions to Overcome Them?
The tool encounters limitations in conducting comprehensive investigations of artworks, particularly concerning surface cracks, color distortion, and canvas elasticity. The specimen may have undergone multiple restorations or may have suffered irreversible damage, which can lead to deviations in its original parameters, complicating the validation of its authenticity. Keras360 has implemented minor adjustments to the user interface to facilitate registered users in uploading a reference image alongside a test image. The AF tool is designed to analyze the similarities between these two images. This feature is expected to operate in a secure domain, distinct from public access.
Keras360 水の家 