New breakthrough in AI replacement: after 1 million photos training, the accuracy of deconstructing and reconstructing clothing is 95.7%.

CTOnews.com, Nov. 28 (Xinhua)-- A research team has recently developed a Sewformer AI system that can identify and reconstruct the costumes of characters in pictures after training 1 million clothing photos.

The Sewformer AI system can analyze the data of a single image entered by the user, deconstruct the clothing worn by the character in the picture, and successfully find out each part and gather it together to reconstruct, with an accuracy of 95.7%, thus making the clothing generated by AI more real.

The researchers say the most important potential use of the system is virtual reality / meta-universe, which can map and construct the current dress of the wearer in the digital world, and expand paid revenue models such as shopping malls.

Of course, there are other possibilities for clothing made by artificial intelligence, such as taking a picture of someone wearing something, then recreating it, and making a new design based on it. At present, research on this experiment is provided on the preprint server arXiv.

CTOnews.com attached the reference address of the paper: https://arxiv.org/abs/2311.04218