Food101Personal, VFN-Personal

本研究介绍了两个个性化食品图像分类的基准数据集：Food101Personal和VFN-Personal。Food101Personal基于真实世界参与者的日常饮食模式调查创建，包含20个模式，每个模式300张图像，平均44个食品类别。VFN-Personal则基于一项饮食研究，包含26个模式，每个模式300张图像，平均29个食品类别。这两个数据集通过模拟长期个性化饮食模式来增强其代表性。数据集的应用领域是解决个性化食品分类问题，通过训练深度神经网络以反映每个个体的饮食模式，提高食品分类的准确性。

This study introduces two benchmark datasets for personalized food image classification: Food101Personal and VFN-Personal. Food101Personal is developed based on surveys of daily dietary patterns from real-world participants, consisting of 20 dietary patterns, with 300 images per pattern and an average of 44 food categories per pattern. VFN-Personal, on the other hand, is based on a dietary study, containing 26 dietary patterns, 300 images per pattern, and an average of 29 food categories per pattern. Both datasets enhance their representativeness by simulating long-term personalized dietary patterns. These datasets are applied to solve personalized food classification tasks, where deep neural networks are trained to reflect the dietary patterns of individual users, thereby improving the accuracy of food classification.