A 1940 President Club sedan, in green, at the Studebaker Museum in South Bend, Indiana

In 2015, documentary photographer Carol Highsmith received a letter from Getty Images accusing her of copyright infringement for featuring one of her own photographs on her own website. It demanded payment of $120. This was how Highsmith came to learn that stock photo agencies Getty and Alamy had been sending similar threat letters and charging fees to users of her images, which she had donated to the Library of Congress for use by the general public at no charge. In 2016, Highsmith has filed a $1 billion copyright infringement suit against both Alamy and Getty stating “gross misuse” of 18,755 of her photographs. “The defendants [Getty Images] have apparently misappropriated Ms. Highsmith’s generous gift to the American people,” the complaint reads. “[They] are not only unlawfully charging licensing fees … but are falsely and fraudulently holding themselves out as the exclusive copyright owner.” According to the lawsuit, Getty and Alamy, on their websites, have been selling licenses for thousands of Highsmith’s photographs, many without her name attached to them and stamped with “false watermarks.” (more: http://hyperallergic.com/314079/photographer-files-1-billion-suit-against-getty-for-licensing-her-public-domain-images/)

We at GetArchive are exploring new methods of image metadata augmentation and verification. Our goal is to make it possible to find images on any topic. In particular, we are trying to verify and fix historic periodization. This collection is made of historic photographs of automobiles that look as if they were taken in the 1940s. The collection is made with aid of a neural image recognition network dealing with the whole image composition rather than with the car model - some cars may be dated incorrectly. Although, while this method is surprisingly good for the purpose of dating and tagging, a certain percentage of images (less than 8%) may not represent automobile, but other vehicle type or visually similar object. Naturally, our next step should be creating numerous datasets for a particular car years&models, but as of September 2022, we found no use to justify the effort.