We present Neural Congealing -- a zero-shot self-supervised framework for detecting and jointly aligning semantically-common content across a given set of images. Our approach harnesses the power of pre-trained DINO-ViT features to learn: (i) a joint semantic atlas -- a 2D grid that captures the mode of DINO-ViT features in the input set, and (ii) dense mappings from the unified atlas to each of the input images. We derive a new robust self-supervised framework that optimizes the atlas representation and mappings per image set, requiring only a few real-world images as input without any additional input information (e.g., segmentation masks). Notably, we design our losses and training paradigm to account only for the shared content under severe variations in appearance, pose, background clutter or other distracting objects. We demonstrate results on a plethora of challenging image sets including sets of mixed domains (e.g., aligning images depicting sculpture and artwork of cats), sets depicting related yet different object categories (e.g., dogs and tigers), or domains for which large-scale training data is scarce (e.g., coffee mugs). We thoroughly evaluate our method and show that our test-time optimization approach performs favorably compared to a state-of-the-art method that requires extensive training on large-scale datasets.

翻译：我们提出神经凝合（Neural Congealing）——一种零样本自监督框架，用于检测并联合对齐一组给定图像中语义共有的内容。我们的方法利用预训练的DINO-ViT特征来学习：（i）联合语义图谱——一个捕获输入集合中DINO-ViT特征模式的二维网格，以及（ii）从统一图谱到每幅输入图像的密集映射。我们推导出一个新的鲁棒自监督框架，该框架针对每幅图像集优化图谱表示和映射，仅需少量真实世界图像作为输入，无需任何额外输入信息（例如分割掩码）。值得注意的是，我们设计的损失函数和训练范式仅考虑外观、姿态、背景杂乱或其他干扰物体剧烈变化下的共享内容。我们在大量具有挑战性的图像集上展示了结果，包括混合域图像集（例如，对齐描绘猫的雕塑和艺术品的图像）、描述相关但不同物体类别（例如狗和老虎）的图像集，或大规模训练数据稀缺的域（例如咖啡杯）。我们全面评估了我们的方法，并表明我们的测试时优化方法在性能上优于需要在大规模数据集上进行广泛训练的最先进方法。