NutritionVerse-Synth
收藏arXiv2023-12-11 更新2024-06-21 收录
下载链接：
https://saeejithnair.github.io/nvsynth
下载链接
链接失效反馈官方服务：
资源简介：
NutritionVerse-Synth（NV-Synth）是由滑铁卢大学创建的大型合成食品图像数据集，包含84,984张高分辨率的2D食品图像，源自7,082个程序生成的3D场景。该数据集通过动态和程序化两种方法生成，确保了食品场景的多样性和真实性。每个场景从12个不同视角捕捉，并附有详尽的营养信息和多模态标注，如RGB、深度、语义分割等。NV-Synth旨在推动计算机视觉技术在自动化饮食评估中的应用，解决传统营养跟踪方法的不足。
NutritionVerse-Synth (NV-Synth) is a large-scale synthetic food image dataset developed by the University of Waterloo. It contains 84,984 high-resolution 2D food images derived from 7,082 procedurally generated 3D scenes. This dataset is generated via both dynamic and procedural methods, ensuring the diversity and realism of food scenes. Each scene is captured from 12 distinct perspectives, and is accompanied by comprehensive nutritional information and multimodal annotations such as RGB data, depth maps, semantic segmentation masks and so on. NV-Synth aims to advance the application of computer vision technologies in automated dietary assessment, addressing the limitations of traditional nutrition tracking methods.
提供机构：
滑铁卢大学
创建时间：
2023-12-11
