The StableDrag drag-and-drop technology jointly developed by Nanjing University and Tencent has been upgraded again, bringing a more stable and accurate image editing experience. The core of this upgrade is to build a more stable drag-and-drop editing framework through discriminative point tracking and confident action supervision methods. Improvements in this technology have significantly improved the quality and consistency of image editing, providing users with a better image processing experience.
The StableDrag drag-and-drop P-picture technology jointly launched by Nanjing University and Tencent has been upgraded again, making it more stable and accurate. The researchers built a stable drag-and-drop editing framework through discriminative point tracking and confident action supervision methods. Experimental results show that StableDrag-GAN and StableDrag-Diff show more accurate and stable editing effects in DragBench evaluation, improving the quality and consistency of image editing.
The upgrade of StableDrag technology highlights the remarkable results achieved by cooperation between academia and industry in the field of artificial intelligence image processing. It is expected to play an important role in more image editing applications in the future, bringing users a more convenient and efficient image processing experience. We look forward to StableDrag bringing more innovations and breakthroughs in the future.