Unsupervised learning of keypoints and landmarks has seen significant progress with the help of modern neural network architectures, but performance is yet to match the supervised counterpart, making their practicability questionable. We leverage the emergent knowledge within text-to-image diffusion models, towards more robust unsupervised keypoints. Our core idea is to find text embeddings that would cause the generative model to consistently attend to compact regions in images (i.e. keypoints). To do so, we simply optimize the text embedding such that the cross-attention maps within the denoising network are localized as Gaussians with small standard deviations. We validate our performance on multiple datasets: the CelebA, CUB-200-2011, Tai-Chi-HD, DeepFashion, and Human3.6m datasets. We achieve significantly improved accuracy, sometimes even outperforming supervised ones, particularly for data that is non-aligned and less curated. Our code is publicly available and can be found through our project page: https://ubc-vision.github.io/StableKeypoints/

利用文本到图像扩散模型的新知识，我们的核心思想是找到使生成模型始终关注图像中的紧凑区域（即关键点）的文本嵌入。通过优化文本嵌入，使去噪网络内的交叉注意力图定位为具有小标准差的高斯分布，我们在多个数据集上验证了性能，包括CelebA、CUB-200-2011、Tai-Chi-HD、DeepFashion和Human3.6m数据集。在非对齐和不太精选的数据上，我们取得了显著改进的准确性，有时甚至超过了监督学习方法。我们的代码公开可用，可以在我们的项目页面找到。

预训练扩散模型的无监督关键点