Egbert van Zwan


Display windows #1, c. 2000, gelatin silver print 19x27 cm  
← previous next →
The computer scientist Egbert van Zwan photographed because he needed work for his scientific research in the field of image recognition. When it appeared that the computer models he developed did not function the way he envisioned, he gave up his scientific project as well as his photographic efforts.

Photography as Scientific Experiment
In 1994 Van Zwan commenced his studies in computer science at the Radboud University in Nijmegen. He was especially interested in artificial intelligence and the handling of uncertain knowledge. The idea came to him for a possible application of methods used in the field of artificial intelligence to the analysis of images. He wrote a pioneering surface-representation program, for which he graduated cum laude in 1999. Using this program, it is possible to determine the probability of specific two-dimensional images in which chance and uncertainty play a role. By chance, one should think especially of photographs because there are always unlikely elements to be found in the image, but the model can also be used for the analysis of certain paintings (for example the ‘action paintings’ of Jackson Pollock). Uncertainty plays a role in computer simulation for which incomplete information is available, such as constructed images of criminals or extinct animals.

The program of Van Zwan lies at the basis of several commercial applications in the field of verification of photographs and extrapolation of incomplete images. Thus, whether a photo has been manipulated can be established from measurements of non-trivial relationships between elements within the image, combined with a statistical comparison of all possible perspective transformations. Particularly in the current digital era, this will become increasingly important. Another application is the determination of the maximum probability in the composition of synthetic portraits. This is not only relevant to the aforementioned profiles of criminals, but also to the composition of prephotographic portraits on the basis of non-photographic information. Through advanced calculations using shape modelling with subdivision surfaces, the program determines whether certain combinations of facial features are possible. The maximum of the probability distribution for each independent element will ultimately generate the most probable picture of the face.

These applications, while important, were only side issues as far as Van Zwan was concerned. During his doctoral research, which he started after graduating in 1999, he became actively interested in the medium of photography. He began to study its history, its theory and started collecting photo books. Using a self-made computer model, he wanted to make a contribution to photographic interpretation and criticism by showing that it is possible, with an arbitrary photograph, to say, with mathematical certainty, something about its maker and his intention. This research touches on fundamental issues of interest in photography related to the question of whether or not the personal intention of the photographer can be communicated by the photographs themselves. The sceptic might argue that, due to the mechanical nature of photography and the automatic relationship that exists between the object in reality and in the picture, it is not possible to communicate a personal intention with respect to that object. In contrast, Van Zwan believed that the choice of subject matter, the framing of the image, its composition and the moment of photographing are intentional acts that do communicate something about the character and intent of the photographer. He imagined a program that would analyse scanned photographs and based on the results, would generate a psychological profile of the photographer and an interpretation of his intentions. As a test for this program, he devised that he would require a statistically significant quantity of work of a particular photographer, one who takes pictures with many random elements in the image. To this end he started his own photographic career. He took pictures in shopping area's because he was interested how his program would react to pictures with both people and window dummies.
All of his images, independently of their quality, were scanned and decomposed with his computer program. Initial results were encouraging. After analyzing hundreds of photographs, a profile was created in which Van Zwan recognized himself. Moreover, the program generated, for each individual picture, an output containing a striking observation about the photographer, which at first sight appeared to be non-trivial. Closer examination, however, showed that the results were not entirely consistent. This was especially evident with very similar photos which were taken in rapid succession, where the ‘comments’ sometimes differed widely from each other. To eliminate these imperfections, Van Zwan adjusted his program at various points. One of the first improvements was the addition of a self-developing component, a feature identification element based on evolutionary strategies. He appeared to be moving in the right direction, nevertheless some anomalies in the results remained. To remedy these, he integrated an autoregressive form model and para-perspective imaging. After numerous refinements and as many less than satisfactory results, it became clear to Van Zwan that, even with such an advanced mathematical method, the accuracy he desired could not be achieved. He realized that the time was not yet ripe for his project. Not because the goal he had set for himself, i.e. to say something meaningful about the photographer and his intentions with an arbitrary photograph, was based on a wrong hypothesis, but because mathematical knowledge was not yet rich enough for such a model.
This realization led to his complete collapse and made him withdraw from public life for years. He gave up science and photography and never finished his Ph.D. research, which, apart from a few formalities, had been completed. Recently, he took a job as a librarian. I bought his entire photo collection and have made a selection therefrom.

Related photographers

Onno Meerdink


Sammy Hegontvlug