Approximately 3,000 top-down images of realistic food plates from Google cafeterias, captured using a custom scanning rig. The dataset provides nutritional annotations including calories, fat, carbohydrate, and protein for the total plate and for individual ingredients. It was created by author sunli1201 and last updated on Hugging Face in April 2026.
Use Cases
- Train food recognition models based on annotated plate images.
- Develop calorie and macronutrient estimation systems based on visual and nutritional data.
- Analyze ingredient-level nutritional composition based on the provided annotations.
- Benchmark multimodal models that combine image and tabular nutritional data.
Strengths
- Approximately 3,000 data points provide a substantive sample size.
- Includes both visual (640x640 images) and detailed nutritional annotations (calories, macronutrients).
- Provides nutritional data at both the total plate and per-ingredient level.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count, file formats, and license are unknown, which may limit suitability assessment.
- Data may reflect geographic and culinary bias inherent to its source (Google cafeterias).
Provenance
- Source
- Google cafeterias, via a custom scanning rig.
- Collection Method
- Images captured from realistic food plates using a custom scanning rig.
- Time Range
- null
- Freshness
- Last updated 2026-04-02 11:08:19; freshness should be verified.
- Geography
- null