EDEN

EDEN数据集是由阿姆斯特丹大学计算机视觉实验室创建的多模态合成数据集，专注于封闭的园艺场景。该数据集包含超过30万张从100多个花园模型捕获的图像，每张图像都标注了多种视觉模态，如语义分割、深度、表面法线、固有颜色和光流。数据集的创建旨在促进农业和园艺等自然导向应用的计算机视觉和机器学习方法的发展。通过使用物理基础渲染器，数据集能够模拟不同光照条件下的场景，从而提供高质量的图像用于训练低级计算机视觉任务。EDEN数据集的应用领域主要集中在解决自然场景中的视觉问题，如语义分割和单目深度预测，这些问题对于农业和园艺机器人的发展至关重要。

The EDEN dataset is a multimodal synthetic dataset created by the Computer Vision Lab at the University of Amsterdam, dedicated to enclosed horticultural scenarios. It contains over 300,000 images captured from more than 100 garden models, with each image annotated with multiple visual modalities including semantic segmentation, depth maps, surface normals, intrinsic colors, and optical flow. The dataset is developed to promote the advancement of computer vision and machine learning methods for nature-oriented applications such as agriculture and horticulture. By leveraging physics-based renderers, it can simulate scenes under diverse lighting conditions, thereby providing high-quality images for training low-level computer vision tasks. The primary application areas of the EDEN dataset focus on addressing visual challenges in natural scenes, such as semantic segmentation and monocular depth prediction—tasks that are critically important for the development of agricultural and horticultural robots.