Photoshop, the famous Adobe software, is 28 years old. Since the 1990s, many Sunday designers have used Adobe Photoshop to retouch photos. The tools offered today by Photoshop make it easy to modify the images of your choice. Although these edits are usually done for good purposes, some may use these software to manipulate opinion, by broadcasting fake photos that do not correspond to reality. Artificial intelligence can now facilitate the detection of retouched photos.
The machine learning to identify retouched photos
Adobe is now using machine learning to easily spot retouched photographs. Original sets of photos and their retouched equivalents have fueled an artificial intelligence, which can now detect if a photo is retouched.
Several methods already allowed to identify retouched photos (metadata, noise distribution, lighting, contrasted edges …). Vlad Morariu, senior research scientist at Adobe, had the good idea to combine these proven techniques with artificial intelligence to create a tool able to estimate if a photo was retouched or not. More precisely, it allows you to see items moved, added and deleted on photos. The democratization of this type of tool, made possible thanks to advances in artificial intelligence associated with specific expertise, could facilitate the detection of retouched photos for questionable purposes.
To learn more about the techniques used by Adobe to identify retouched photos and the researcher’s work, you can watch the video above and view this publication on the Adobe blog.
Eli Washington is still early into his career as a marketing and tech reporter but has already had his worked published in many major publications including Gizmodo and Endgadget. In regards to academics, Eli earned a degree in business from the University of Southern California. Eli follows the money and covers all aspects of search and marketing here at The Marketing Folk.