Evaluating Art with AI: Is it Practical?

Most of the paintings (we will discuss them today) were written decades or even centuries ago. The most prominent of them adorn world-class museums and exhibitions, and some have even found their place in the estates of wealthy citizens.?

Artists' works may cost a lot of money, as can their high-quality copies. However, to determine whether you are looking at an original, a copy, or even a fake, you will need the help of an expert.?

What if I told you that art evaluation can be almost immediate, as well as its search, purchase, or sale?

A recent AdvantISS team project confirms that AI can serve as a tool for generating and verifying information of all types. This is especially true for modern solutions such as ChatGPT 4o, which determines the content quite accurately and can even analyze it quickly.?

So, let's consider whether is it practical to evaluate art with the help of artificial intelligence.

Art Assessment: Bottlenecks of the Process

Each picture, whether original or copied, has its own “code”. It is formed from dozens or even hundreds of different special elements present in the work of art. They determine not only the quality of the painting, but also its value at auction.

However, the modern evaluation system has significant shortcomings, in particular:

• The need for physical contact between the expert and the artwork.
• Numerous parameters should be considered to determine the originality and quality of a painting.
• Impossibility of remote evaluation, for example, through a smartphone camera, etc.

However, if we look at the evaluation process from a different perspective and also involve AI, we can eliminate challenges and shortcomings, significantly speed up the work, and make it easier for connoisseurs to access works of art.?

Evaluation Process and Parameters

In practice, appraising works of art by a specialist can take hours, or even days, or even longer. And this causes many inconveniences for those who want to buy paintings. We have implemented the following system in our platform:

• Full digitization of an artwork. Using serial photography, 3D scanning, and additional tools, we transfer the picture to the information space.
• Analysis and continuous learning. The built-in AI evaluates the work based on hundreds of parameters and compares it in detail with the original (also digitized).
• Structuring artworks. We combine paintings by various factors to make it easier for customers to search and logically place works on the platform.
• Determining the value of the picture. Guided by the best global practices and expert advice, we have developed an algorithm that considers market prices and objectively evaluates available works.?
• Selling art through the platform. We have implemented payment models that allow you to purchase a physical painting or its digital copy in the form of NFT in several ways.?

This way, we minimize the waiting time for the painting to be evaluated and speed up its sale to a new happy owner. By the way, the client is satisfied with the platform, in particular, because of the high accuracy of art valuation based on numerous key parameters.

Image

External surface assessment of the picture. An artificial intelligence bot scans the image with a camera and compares it to the digitized original. Next, algorithms are used to analyze every inch of the graphic in detail, looking for deviations from the standard.

If no deviations are detected, the artwork is assigned the maximum quality level according to the first parameter. In the case of poor quality, the painting is re-identified and assigned a price, but using slightly different algorithms.

Detailing

Every handmade thing has its own characteristics and is always different from a similar work by the same author. Works of art are much more complicated, as they are usually created with a brush and paint. It means that an artist can make a mistake or deliberately change the strokes to add their own details to the painting.

Artificial intelligence uses powerful scanning tools to analyze even subtle details. Then, the algorithm compares them with the original fragments and classifies and re-evaluates them according to additional parameters.?

Texture

If you think that paintings are 100% flat (2D), you are quite wrong. When paint is applied to the canvas, it forms a base layer with which the artist interacts later. The more strokes the artist makes, the more complex the structure becomes and the more voluminous the painting is. Although it is not very noticeable from the outside, the value of a work of art depends on how well it is executed.

The artificial intelligence in the platform has access to state-of-the-art equipment that can take numerous images from different angles. This makes it possible to estimate (approximately) how many layers of paint are applied to the canvas, what technique is used, and other work details.?

Material

Usually, one of several basic materials is applied to create paintings, including paper or fabric canvas. It determines the durability of the artwork, the speed of paint fading, peeling, etc. The material of the frame is also important, but nowadays, they are either restored or new ones are created, so it is not a very critical factor.

Specially trained artificial intelligence uses scanning and viewing to determine the type of material of the painting's base, its quality, and potential durability, adding a few more points to the overall score.

Nuances

Other important aspects of the evaluation include the painting's authorship, date of creation, artistic value, and accompanying information such as title or description. While such data is usually contained in a digitized file, sometimes it is necessary to look for confirmation or compare styles to verify information.?

In such cases, artificial intelligence unleashes its maximum potential and establishes all the facts in seconds, while filling in the artwork profile and generating the appropriate description for the auction page.

Art Historians VS AI

Technically, with artificial intelligence, you can quickly and accurately evaluate paintings, identify their authors, and determine their originality. However, due to hardware limitations and imperfections in the technology, it can also make mistakes.?

AI-assisted valuation is more profitable, primarily because of:

• Information processing speed.
• Accuracy of detail identification.
• Correct definition of materials, colors, etc.

At the same time, technology cannot objectively evaluate works of art and establish their real value. Therefore, the experience of art historians has an advantage in this matter.

As for the practicality of evaluating paintings exclusively by AI algorithms, it still makes sense to ensure the evaluation is correct with the help of experts, but this need will soon disappear.

Do you agree with me? Share your vision of the situation in the comments!