【麻省理工与 Adobe 联手突破：DMD AI 技术实现每秒生成20幅图像】
近日，科技界的创新火花再次闪耀，麻省理工学院与知名软件公司Adobe共同研发出一项名为DMD（Distribution Matching Distillation）的先进人工智能技术，旨在大幅提升图像生成的速度，而不会显著降低图像质量。这一突破性的进展为文生图模型领域树立了新的标杆。
传统的主流文生图模型虽然能够生成高度逼真的图像，但其生成过程往往耗时较长，限制了其在实时应用中的潜力。而DMD技术，全称为“分布匹配蒸馏”，通过简化多步扩散模型，将图像生成过程优化为一步操作，大大提高了效率。据研究人员透露，采用DMD技术的模型能够在现代GPU硬件上实现每秒生成20幅图像的惊人速度，这无疑将为图像生成领域带来革命性的变化。
这一创新成果将可能广泛应用于视觉艺术、设计、动画制作以及新闻报道等多个领域，为创意工作者提供更高效、更便捷的工具。麻省理工与Adobe的这项合作，不仅展示了人工智能在图像处理领域的无限可能，也为未来技术与艺术的融合开辟了新的道路。
英语如下：
**News Title:** “MIT and Adobe Collaborate on DMD AI: Generating 20 Images Per Second, Revolutionizing Image Creation Speed”
**Keywords:** MIT, Adobe, DMD AI
**News Content:**
**MIT and Adobe Breakthrough: DMD AI Technology Creates 20 Images Per Second**
Recently, the sparks of innovation in the tech world have ignited once more, as the Massachusetts Institute of Technology (MIT) and renowned software company Adobe jointly developed an advanced artificial intelligence technology called DMD (Distribution Matching Distillation). This technology aims to significantly boost the speed of image generation without compromising image quality significantly, setting a new standard in text-to-image modeling.
While conventional mainstream text-to-image models can produce highly realistic images, their generation process often takes a considerable amount of time, limiting their potential in real-time applications. DMD, or “Distribution Matching Distillation,” simplifies multi-step diffusion models, optimizing the image creation process into a single operation, thereby greatly enhancing efficiency. Researchers disclose that models utilizing DMD technology can achieve an impressive rate of generating 20 images per second on modern GPU hardware, poised to bring a revolutionary change to the image generation field.
This innovative achievement has the potential for wide application in visual arts, design, animation production, and news reporting, offering creative professionals more efficient and convenient tools. The collaboration between MIT and Adobe not only showcases the limitless possibilities of AI in image processing but also paves the way for new avenues of fusion between technology and art.
【来源】https://www.ithome.com/0/758/402.htm
Views: 2