More

    AI recreates clothing with 95.7% accuracy in breakthrough experiment

    Published on:

    The hype around AI continues to grow, and we are seeing new ways in which it can make our lives easier and make us even more productive. A recent experiment with some success showed that the AI ​​could recreate the garment by simply looking at a single image of the garment in question.

    This was an interesting experiment, and the researchers say they had a 95.7 percent success rate in reproducing the clothes shown in the experiment. This experiment is his two-stage AI system known as Sewformer. The researchers trained the program using 1 million images of her in clothing.

    Its training because the system can look at the image once and basically break down the garment to figure out where the different parts come together and what patterns should be used to recreate it. All seem to have been successful. This could make AI-made clothing more realistic than any previous experiment.

    Image source: Dickies/Amazon

    Of course, one of the most important uses for this type of system is through virtual reality, or the metaverse, the researchers say. This will allow developers to use this technology to recreate digital versions of clothing and place them in the digital world. This will allow users to recreate real-world clothing within the Metaverse, and will allow developers and clothing creators to charge for their items in the digital world.

    Of course, there are other possibilities for AI-created clothing, such as taking a photo of someone wearing something and recreating it to create a new design.a studying experiments Currently available on preprint servers arXiv.

    It’s unclear exactly where the researchers plan to expand this experiment in the future, or whether the Sewformer system will see any widespread adoption within the garment industry. Still, it’s truly amazing to see what AI is capable of when given the right training and tools.

    Related

    Leave a Reply

    Please enter your comment!
    Please enter your name here