arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2212.03099cs.CVcs.CLcs.MM
收藏

语义条件扩散网络在图像描述中的应用

Semantic-Conditional Diffusion Networks for Image Captioning

Sun Yat-sen University(中山大学);JD Explore Academy(京东探索研究院)

Jianjie Luo, Yehao Li, Yingwei Pan, Ting Yao, Jianlin Feng, Hongyang Chao, Tao Mei

摘要:

文本到图像生成的最新进展见证了扩散模型的兴起,它作为强大的生成模型。然而,利用这样的潜变量模型来捕获离散词之间的依赖性,同时在图像描述中追求复杂的视觉语言对齐并不是微不足道的。在本文中,我们打破根深蒂固的公约,学习transformer为基础的编码器-解码器,并提出了一个新的扩散模型为基础的范例量身定制的图像描述,即语义条件扩散网络(SCD-网)。在技术上,对于每幅输入图像,我们首先通过跨模态检索模型搜索语义相关的句子,以传达全面的语义信息。丰富的语义被进一步视为语义优先于触发扩散Transformer的学习,其在扩散过程中产生输出句子。在SCD网络中,多个扩散Transformer结构被堆叠,以级联方式逐步加强输出句子,具有更好的视觉语言对齐和语言连贯性。此外,为了稳定扩散过程,设计了一种新的自临界序列训练策略,以指导SCD网络的学习与标准自回归Transformer模型的知识。COCO数据集上的大量实验表明,在具有挑战性的图像描述任务中使用扩散模型具有很大的潜力。源代码可从\url{https://github.com/YehLi/xmodaler/tree/master/configs/image_caption/scdnet}获得。

英文摘要:

Recent advances on text-to-image generation have witnessed the rise of diffusion models which act as powerful generative models. Nevertheless, it is not trivial to exploit such latent variable models to capture the dependency among discrete words and meanwhile pursue complex visual-language alignment in image captioning. In this paper, we break the deeply rooted conventions in learning Transformer-based encoder-decoder, and propose a new diffusion model based paradigm tailored for image captioning, namely Semantic-Conditional Diffusion Networks (SCD-Net). Technically, for each input image, we first search the semantically relevant sentences via cross-modal retrieval model to convey the comprehensive semantic information. The rich semantics are further regarded as semantic prior to trigger the learning of Diffusion Transformer, which produces the output sentence in a diffusion process. In SCD-Net, multiple Diffusion Transformer structures are stacked to progressively strengthen the output sentence with better visional-language alignment and linguistical coherence in a cascaded manner. Furthermore, to stabilize the diffusion process, a new self-critical sequence training strategy is designed to guide the learning of SCD-Net with the knowledge of a standard autoregressive Transformer model. Extensive experiments on COCO dataset demonstrate the promising potential of using diffusion models in the challenging image captioning task. Source code is available at \url{https://github.com/YehLi/xmodaler/tree/master/configs/image_caption/scdnet}.

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。

↑