The Lost Arles sketchbook

Step by step overview of our methodology for assessing the authenticity of a sketchbook.

Description and history of the sketchbook

is The Lost Arles Sketchbook authentic?

In this article I am assessing the authenticity of the sketchbook published in the book:

The Lost Arles Sketchbook , Welsh- Ovacharov, Bogomila ISBN 10: 1419725947 – ISBN 13: 9781419725944

This sketchbook was promoted as \”The most revolutionary discovery in the entire history of Van Gogh\’s oeuvre. Not one drawing; not ten, not fifty, but sixty-five drawings.\”  by Ronald Pickvance, from the Foreword

Story behind the sketchbook: Late in life, during his time living in Provence, Vincent van Gogh kept a sketchbook within a humble account ledger given to him by Joseph and Marie Ginoux, the owners of the Café de la Gare in Arles. This artifact of incalculable historical and aesthetic value remained hidden for more than one hundred and twenty years. It reappears today as a revelation and an extraordinary treasure.

Rejections from Van Gogh Museum: The sketchbook was rejected as fake by the van Gogh Museum in 2008, 2012, and after the publication of the book in November 2016. They also cite that their opinion was not included in the book. In the last occasion, the VGM press release was very descriptive about the reasons for the rejection, going into detail even for the stylistical aspects. The experts from the museum claim that there are distinctive topological errors and that the discolour in the drawings is uneven accross the artworks, and artificially created. They provide evidences of this chromatic unconsistencies for paper and ink. Additionally, they point out that the drawings do not show the evolution made van Van Gogh during these years, and they are \”monotonous, clumsy and spiritless\”. Also they do not see the characteristic refinement of the Van Gogh drawings.

After reading the VGM press release, the answer from the authors who published the book was very short and vague: \”We have not changed our minds and are very happy that from now on everybody can make their own opinion after seeing the drawings and reading the analysis in the book,\”. \”When I know something is a van Gogh, I know it is a van Gogh,\”

Goal of the study and approach: Our goal is to test our method with all the 65 drawings in the sketchbook. For this, I will train my model using only artworks which are proven to be made by Van Gogh. The artificial networks will not use any other image, in order to preserve the bias-free classification. We will see what the model says about the sketchbook.

PREPARING THE DATASETS

Narrowing it around the Arles period

Selected artworks: van Gogh spent fifteen month living in Arles, from February 1888 to May 1889. In order to train our model on a sample of artworks which closely represents the drawing style of the artist during the mentioned period, we selected the artworks from his 4 latter years of life (1887-1890). For the testing artworks, in addition to the one featured in the referred paper, we reserved some authentic ones, and added some forged drawings introduced by Otto Wacker.

We trained with 184 high resolution images of drawings made by Van Gogh during his time in Arles. We used the De La Faille catalog raisonne to verify the authenticity and provenance of the drawings used for training.

Image preparation: We used our own algorithm to filter the texture and tint of the paper of the drawings. This prevents the network to learn the paper, or even the camera, and to focus the learning on the drawing style. Next, in order to remove the influence of fading or degradation of the colors, we converted the photos of signatures into squared grayscale images.

PREPARING THE DATASETS

Narrowing it around the Arles period

After the image preparation, we chopped the images to use as input for our neural network tuned for drawings, which works with sub-images of 256×256 pixels.

In order to chop the drawings, we used our algorithm which selects the most relevant regions and groups them to improve the performance of the model. The datasets are split as follows: 70% of the authentic signatures were randomly separated for training, 10% for validation, and the remaining 20% were reserved for testing.

On the left hand image, you can see the image with the squared regions selected. You can see that only the regions of the artwork with the largest concentration of strokes are included in the squares, and grouped, at different scales. Each one of these selected regions will be tested for authenticity and studied separately. Additionally, the same is done with the image containing the whole drawing.

Training our neural network

Our neural networks are trained on the images of authentic Van Gogh drawings

The number of available authentic drawings is finite, so we performed data augmentation with the images: This means randomly translating and rotating the patches in the squared images. This procedure increments the variety of inputs to the network, free of biases, and results in an improvement in the accuracy of the classification. Additionally, it is also zoomed in and out a bit.

• We trained our model using only the authentic samples. For this task, we used two networks, an encoder and a decoder, which is able to reconstruct the signatures. After obtaining training convergence, we saved the trained model and tried it with the test signatures to evaluate the performance.

the veredict of our model

Authenticity of The Lost Arles Sketchbook

We cannot reproduce here the individual drawings, due to copyright issues. However we reference the drawings in the book, following by the classification probability of being non authentic.

REFERENCE

PROBABILITY OF BEING NON AUTHENTIC

BW01	89%
BW02	84%
BW03	91%
BW04	88%
BW05	89%
BW06	82%
BW07	87%
BW08	86%
BW09	79%
BW10	84%
BW11	93%
BW12	82%
BW13	80%
BW14	89%
BW15	79%
BW16	96%
BW17	88%
BW18	96%
BW19	75%
BW20	84%
BW21	79%
BW22	72%
BW23	91%
BW24	67%
BW25	89%
BW26	86%
BW27	95%
BW28	93%
BW29	85%
BW30	90%
BW31	93%
BW32	82%
BW33	86%
BW34	89%
BW35	79%
BW36	89%
BW37	82%
BW38	92%
BW39	79%
BW40	82%
BW41	89%
BW42	82%
BW43	87%
BW44	89%
BW45	92%
BW46	82%
BW47	87%
BW48	79%
BW49	82%
BW50	72%
BW51	93%
BW52	82%
BW53	70%
BW54	89%
BW55	79%
BW56	69%
BW57	79%
BW58	89%
BW59	82%
BW60	92%
BW61	79%
BW62	82%
BW63	79%
BW64	94%
BW65	89%

All of them has been detected as forgery by our model. Therefore

The algorithm classifies the sketchbook as Non Authentic.

Testing accuracy: Using our trained model we obtained a prediction accuracy of 89% for the authenticity of 220 drawings. And the average prediction probability of The Lost Arles drawings is 85% of being fake. We consider this confidence enough to issue a certificate with the detailed description of the experiment.

Final remarks: Our model agrees with the decision of the Van Gogh Museum to reject the sketchbook as non authentic. Our model was able to corroborate the non-authenticity of the artworks, using as evidence solely a comparision of the style of the drawings in the sketchbook with the style of drawings from Van Gogh.

Need help to authenticate or value an item?

We offer a personalized service to our customers.

Four Easy Steps to Request Your Item\’s True Value:

Error: Contact form not found.

If you find difficulties uploading the images, send them directly to [email protected] .