New AI model fills in blank spots in photos
Japanese university's tech can also slightly modify facial features
HIROKO FUJII, Nikkei staff writer
TOKYO -- A new technology uses artificial intelligence to generate synthetic images that can pass as real.
The technology was developed by a team led by Hiroshi Ishikawa, a professor at Japan's Waseda University. It uses convolutional neural networks, a type of deep learning, to predict missing parts of images. The technology could be used in photo-editing apps. It can also be used to generate 3-D images from real 2-D images.
The team at first prepared some 8 million images of real landscapes, human faces and other subjects. Using special software, the team generated numerous versions for each image, randomly adding artificial blanks of various shapes, sizes and positions. With all the data, the model took three months to learn how to predict the blanks so that it could fill them in and make the resultant images look identical to the originals.
The model's learning algorithm first predicts and fills in blanks. It then evaluates how consistent the added part is with its surroundings.
Consistency is examined around the edges of the real and fake parts. The model then judges whether the entire picture looks natural and real. Ishikawa says the real innovation lays in the technology's ability.
After repeating the cycle of prediction and evaluation, the learning model becomes capable of artificially building up an entire image by using only predicted parts. This allows it to take a real 2-D image and make it 3-D.
The team expects its model to be used in photo-editing apps. The system is also capable of slightly modifying images, like those of children's facial features. The team thinks this will allow parents to lightly doctor pictures of their kids, then post the images on social media platforms with little worry of putting their identities at risk.