Kees van Dongen, “Nu en buste”

Kees van Dongen, a Dutch-French painter, was a leading figure among the Fauves. His early work reflected influences from the Hague School and Symbolism, gradually evolving into a rough pointillist style.
Two versions. Over the past decades, two nearly identical versions of a painting titled “Nu en Buste”, attributed to Kees van Dongen, have surfaced at auction.
Version A was sold at auction in 2008, accompanied by a certificate of authenticity from the Wildenstein Institute. However, when the painting reappeared for auction in Paris in December 2019, the sale was abruptly canceled. The Wildenstein Plattner Institute (WPI) rejected the original appraisal by the Wildenstein Institute and refused to include the piece in the Catalogue Raisonné. WPI declined to explain their decision.
In May 2020, Version B was auctioned in Paris, this time with a certificate from the WPI. The institute indicated a willingness to reconsider their verdict on Version A, provided the owner could supply provenance records. Unfortunately, the owner could only cite the 2008 auction purchase and a mention in a 1989 catalogue. The auction house responsible for the 2008 sale also failed to offer further details on the painting’s provenance.
Amid this complex scenario, the owner sought clarity by turning to Art Recognition.


The Art Recognition AI System in Action. After receiving high-resolution images of the two paintings, we began by training our AI system to identify the defining characteristics of Kees van Dongen’s style. For this, we compiled a dataset of images from van Dongen’s original works, referencing paintings listed in exhibition catalogues such as Musée National d’Art Moderne, Paris (1967), Museum Boymans-van Beuningen, Amsterdam (1968), and Fondation Pierre Gianadda, Martigny (2002). To enhance the AI’s ability to distinguish between authentic works and imitations, we also included paintings by contemporaneous artists with similar styles, such as Otto Müller and Albert Marquet.
Conclusion. Once the AI system was trained to recognize van Dongen’s unique style, we analyzed the two paintings. The results were striking: the AI classified Version A as a forgery with a 73% probability, while Version B was identified as authentic with an 81% probability.
We also conducted a detailed patch-by-patch analysis of the paintings. This involved dividing the images into smaller sections and evaluating the authenticity of each. For Version A, the patches deemed least authentic included the breasts (91% and 87% not authentic), the hair (75% not authentic), and the upper part of the face (85% not authentic).
Ultimately, Version B was sold at auction, while Version A remains a remarkable imitation.