Food2K 大型食品识别数据集

Food2K 是一个大型食品识别数据集，包含 2,000 个类别和超过 100 万张图片。与现有的食品识别数据集相比，Food2K 在类别和图像方面都比现有数据集高出一个数量级，从而为开发用于食品视觉表征学习的高级模型建立了一个新的具有挑战性的基准。可以进一步探索 Food2K，以造福更多与食品相关的任务，包括新兴和更复杂的任务（例如，对食品的营养理解），并且可以预期在 Food2K 上训练的模型可以作为主干，以提高更多与食品相关的任务的性能。

Food2K is a large-scale food recognition dataset encompassing 2,000 categories and over 1 million images. Compared with existing food recognition datasets, Food2K boasts an order of magnitude more categories and image samples, thereby establishing a novel and challenging benchmark for developing advanced models tailored for food visual representation learning. Food2K can be further leveraged to support a broader array of food-related tasks, including emerging and more complex ones such as nutritional understanding of food. It is anticipated that models trained on Food2K can serve as backbone networks to enhance the performance of more food-related downstream tasks.