issai/Food_Portion_Benchmark

Food Portion Benchmark (FPB)数据集是一个包含食物检测和份量估计的多任务食物场景理解数据集，适用于饮食分析、营养跟踪和食物计算研究。数据集包含14,083张图片，涵盖138个食物类别，并提供了边界框和真实重量注释。图片包括顶视图和四个侧面视图，使用Intel RealSense D455和智能手机拍摄。数据集分为训练集、验证集和测试集，并在受控的实验室环境中使用中亚本地菜肴进行收集。

The Food Portion Benchmark (FPB) dataset is a multi-task food scene understanding dataset and benchmark suite that combines food detection and portion size (weight) estimation, suitable for research in dietary analysis, nutrition tracking, and food computing. It contains 14,083 images across 138 food classes, with bounding box and ground-truth weight annotations. The images include top-down and four side views, captured using Intel RealSense D455 and smartphones. The dataset is split into training, validation, and test sets, collected in a controlled lab environment with local Central Asian cuisine.