<ul class="dashed" data-apple-notes-indent-amount="0"><li><span style="font-family: '.PingFangUITextSC-Regular'">文章标题：</span>DreamMatcher: Appearance Matching Self-Attention for Semantically-Consistent Text-to-Image Personalization</li><li><span style="font-family: '.PingFangSC-Regular'">文章地址：</span><a href="https://arxiv.org/abs/2402.09812">https://arxiv.org/abs/2402.09812</a> </li><li>CVPR 2024</li></ul> <img src="https://res.cloudinary.com/montaigne-io/image/upload/v1734959839/C174856F-E217-4384-B0B3-B4F9FB6D9899.png" style="background-color:initial;max-width:min(100%,2382px);max-height:min(948px);;background-image:url(https://res.cloudinary.com/montaigne-io/image/upload/v1734959839/C174856F-E217-4384-B0B3-B4F9FB6D9899.png);height:auto;width:100%;object-fit:cover;background-size:cover;display:block;" width="2382" height="948"><ul class="dashed" data-apple-notes-indent-amount="0"><li></li></ul> 该文章是在定制化生成模型基础上增强生成质量的方法。 具体来说先将参考图像进行DDIM反转得到其扩散过程，然后在目标图像生成的过程中，将参考过程Self-Attention中的Value经过所谓的语义配对变形后替换到目标过程的value中，这里涉及到变形场的概念，需要学一下，大致就是 <img src="https://res.cloudinary.com/montaigne-io/image/upload/v1734959839/7C13D354-2EF7-4FB0-94D3-16BDBEE5CFBD.png" style="background-color:initial;max-width:min(100%,2352px);max-height:min(1404px);;background-image:url(https://res.cloudinary.com/montaigne-io/image/upload/v1734959839/7C13D354-2EF7-4FB0-94D3-16BDBEE5CFBD.png);height:auto;width:100%;object-fit:cover;background-size:cover;display:block;" width="2352" height="1404">
