Art Forensics and Many Others
Time to read: 2 minutes
In December 2024, a comprehensive beta experiment was initiated on the Keras360.io Art Forensics (AF) website. The primary objective was to first seek for matched image(s) in the AF database , and subsequently facilitate further inspection for authentication purposes. However, it has become evident that during segmentation analysis, commonly known as cropped images, the results may lack specificity, leading to diminished accuracy. This observation holds true in the context of zoomed or partial image search protocols.
On August 11th, 2025, we have introduced a new protocol designed to assist in the identification of match images and to verify whether the specified crop upload is derived from the original artwork.
Example utilized in this AF experiment
AF library: Rembrandt. Number 12 (Rembrandt – 12)
Original title: Syndics of the Drapers’ Guild
(Dutch: De waardijns van het Amsterdamse lakenbereidersgilde)
Segmentation 1 Search (Previous)
In the previous protocol, cropped images may be uploaded in either of the two designated fields: i) the full image upload (first from the top), or ii) the zoomed or partial image utilizing the Dice Coefficient method. In instances where no suitable image is identified, the AF beta system will recommend the top three artists and potential artworks. However, for purposes of forensics and authentication, which necessitate a binary response—”Yes” or “No”—the AF beta protocol requires additional input to enhance its effectiveness. Please refer to the video below for details regarding the new protocol; Segmentation 2 Search.
Segmentation 2 Search (Current, New!)
In a manner akin to the prior experiment, a grayscale image has been uploaded to identify its corresponding match within the AF database library. This approach is particularly advantageous when artworks or artefacts exhibit degradation in their original hues and well-defined outlines. Consequently, this method has been retained. The new protocol now mandates an additional upload following a comprehensive image analysis. Users are required to submit a complete image for the search, upon which a pop-up prompt from the User Interface (UI) will request further permission to proceed with segmentation analysis. The resulting output provides a more reliable answer compared to a standard segmentation search.
♤
For enquiries: contact.general@keras360.io
Relevant pages:
All rights reserved – ©The Waterhouse 水の家.
Keras360 水の家 