Meal images and carbohydrate content pr. meal
Description
The dataset comprises 240 images of meals, fruits, and vegetables, with the composed meals representing common Nordic cuisine. The collection includes 120 images without a size reference and 120 images with a size reference included in the frame, where the size reference is a student ID or library card. Consequently, the dataset contains 60 images of composed meals and 60 images of various fruits and vegetables, with each image having an identical counterpart that includes the size reference. All images are labeled with a number from 1 to 60 for fruits and vegetables and 1 to 60 for composed meals. The images were captured by the four authors, each contributing 60 images, taken from a bird's-eye perspective to ensure the food items are clearly visible and properly framed. All images were shot using smartphones equipped with cameras of 12 megapixels or higher, maintaining consistent angles and distances for optimal clarity. A consistent setting with a uniform background and natural lighting to minimize external variables were used. The ingredients for the meals, as well as the fruits and vegetables, were all purchased from Danish supermarkets. Additionally, an accompanying dataset provides the calculated carbohydrate content of the meals, fruits, and vegetables, which can serve as a baseline value. Users of this dataset should be aware that the images are not 100% representative of real-world conditions. To provide a detailed characterization of the composite meals, these are defined as meals consisting of a combination of mixed ingredients, where in many cases the individual components cannot be easily separated visually due to blending, layering, or integration during preparation (e.g., stews, or soups where ingredients are intermixed). However, some composite meals feature components that remain visually distinct and separable (e.g., plated dishes with side items like proteins, grains, and vegetables arranged separately on the plate). No strict rule was enforced requiring the separation of components; instead, the meals were prepared to reflect typical real-world presentations, allowing for natural variability in how ingredients are combined or displayed. This approach ensures the dataset captures a realistic spectrum of meal compositions, from fully integrated mixtures to those with discernible elements. For fruits and vegetables only the edible portions are depicted in terms of quantity, aligning with the actual weight typically consumed by individuals (excluding non-edible parts such as cores, stems, pits, or peels).
Files
Institutions
- Aalborg Universitet