No image is more closely associated with the birth of the motion picture than a train pulling into the French coastal town of La Ciotat. Captured by cinema pioneers Auguste and Louis Lumière, the 50-second clip frightened the audience at its first screening in 1896, who thought a real locomotive was hurtling toward them — or so the legend goes. Those early viewers may simply have felt a technological astonishment we can no longer muster today, and certainly not in response to such a mundane sight. That goes double for the slightly shorter and older Lumière Brothers production La Sortie de l'Usine Lumière a Lyon. Though it depicts nothing more than workers leaving a factory at the end of the day, it has long been referred to as "the first real motion picture ever made."
That qualifier "real," of course, hints at the existence of a predecessor. Whereas La Sortie de l'Usine Lumière a Lyon premiered in 1895, Louis Le Prince's Roundhay Garden Scene dates to 1888. With its runtime under two seconds, this depiction of a moment in the life of four figures, a younger man and woman and an older man and woman, would even by the standards of the Lumière Brothers' day barely count as a movie at all.
Equally disqualifying is its low frame rate: just seven to twelve per second (which one it is has been a matter of some dispute), which strikes our eyes more as a rapid sequence of still photographs than as continuous motion. Even so, it must have been a thrill of a result for Le Prince, an England-based French artist-inventor who had been developing his motion-photography system in secrecy since early in the decade.
We now have a clearer sense of the action captured in Roundhay Garden Scene thanks to the efforts Youtube-based film restorationist Denis Shiryaev, who's used neural networks to bring the historic film more fully to life. Taking a scan of Le Prince's original paper film, Shiryaev "manually cut this scan into individual frames and centered each image in the frame," he says in the video at the top of the post. He then "added a stabilization algorithm and applied an aggressive face recognition neural network in order to add more details to the faces." There followed adjustments for consistency in brightness, damage repairs, and the work of "an ensemble of neural networks" to upscale the footage to as high a resolution as possible, interpolating as many frames as possible. We may feel startled by the lifelike quality of the result in much the same way as 19th-century viewers by the Lumière Brothers' train — which, as we've previously featured here on Open Culture, has also received the Shiryaev treatment.
Based in Seoul, Colin Marshall writes and broadcasts on cities, language, and culture. His projects include the book The Stateless City: a Walk through 21st-Century Los Angeles and the video series The City in Cinema. Follow him on Twitter at @colinmarshall, on Facebook, or on Instagram.