Comprehensive Human Olfactory Perception Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/pyrfume/pyrfume-data
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了8,503个分子与118个气味描述符的配对信息，这些数据来自多个专家标注的来源。数据集在分子与气味描述符的数量上存在不平衡，反之亦然。为了解决这一问题，数据集已经过清洗，合并了重叠部分并过滤掉了罕见描述符。规模上，数据集涵盖了8,503个分子和118个气味描述符。该数据集的任务是预测人类嗅觉感知从分子结构出发。
This dataset contains paired information of 8,503 molecules and 118 odor descriptors, sourced from multiple expert-annotated resources. There is an imbalance in the quantity distribution between molecules and odor descriptors, and vice versa. To address this issue, the dataset has undergone cleaning procedures, including merging overlapping entries and filtering out rare descriptors. The cleaned dataset covers 8,503 molecules and 118 odor descriptors. The core task of this dataset is to predict human olfactory perception based on molecular structures.
提供机构：
Various expert-labeled sources including Arctander’s dataset, AromaDb, FlavorDb, and more.
