arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2401.02015cs.CVcs.AIcs.LG
收藏

利用上下文预测改进基于扩散的图像合成

Improving Diffusion-Based Image Synthesis with Context Prediction

Peking University(北京大学);Tsinghua University(清华大学)

Ling Yang, Jingwei Liu, Shenda Hong, Zhilong Zhang, Zhilin Huang, Zheming Cai, Wentao Zhang, Bin Cui

AI总结:

提出ConPreDiff框架,通过在训练阶段引入上下文预测使每个点保留邻域语义连接,从而改进基于扩散的图像合成,且在推理时不增加参数,在多项图像生成任务上达到SOTA。

摘要:

扩散模型是一类新型生成模型,以前所未有的质量和多样性极大地推动了图像生成的发展。现有扩散模型主要试图沿空间轴通过像素级或特征级约束从损坏的图像中重建输入图像。然而,这种基于点的重建可能无法使每个预测的像素/特征充分保留其邻域上下文,从而损害基于扩散的图像合成。作为自动监督信号的强大来源,上下文在表示学习方面已被深入研究。受此启发,我们首次提出ConPreDiff,通过上下文预测来改进基于扩散的图像合成。在训练阶段,我们显式地强制每个点预测其邻域上下文(即多步长特征/标记/像素),并在扩散去噪块的末端使用上下文解码器,在推理时移除该解码器。这样,每个点可以通过保留其与邻域上下文的语义连接来更好地重建自身。ConPreDiff的这一新范式可以推广到任意的离散和连续扩散主干,且在采样过程中不引入额外参数。我们在无条件图像生成、文本到图像生成和图像修复任务上进行了大量实验。我们的ConPreDiff始终优于先前方法,并在MS-COCO上取得了新的SOTA文本到图像生成结果,零样本FID分数为6.21。

英文摘要:

Diffusion models are a new class of generative models, and have dramatically promoted image generation with unprecedented quality and diversity. Existing diffusion models mainly try to reconstruct input image from a corrupted one with a pixel-wise or feature-wise constraint along spatial axes. However, such point-based reconstruction may fail to make each predicted pixel/feature fully preserve its neighborhood context, impairing diffusion-based image synthesis. As a powerful source of automatic supervisory signal, context has been well studied for learning representations. Inspired by this, we for the first time propose ConPreDiff to improve diffusion-based image synthesis with context prediction. We explicitly reinforce each point to predict its neighborhood context (i.e., multi-stride features/tokens/pixels) with a context decoder at the end of diffusion denoising blocks in training stage, and remove the decoder for inference. In this way, each point can better reconstruct itself by preserving its semantic connections with neighborhood context. This new paradigm of ConPreDiff can generalize to arbitrary discrete and continuous diffusion backbones without introducing extra parameters in sampling procedure. Extensive experiments are conducted on unconditional image generation, text-to-image generation and image inpainting tasks. Our ConPreDiff consistently outperforms previous methods and achieves a new SOTA text-to-image generation results on MS-COCO, with a zero-shot FID score of 6.21.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑