A Deep Learning-Based Multi-Food Detection and Portion Estimation Framework for Automated Nutritional Analysis
Accurate food recognition and nutritional analysis are important components of digital health and dietary monitoring applications. Existing image-based food recognition systems commonly focus on identifying a single food category from an uploaded image, which limits their effectiveness when analyzing meals containing multiple food items. Furthermore, recognizing a food category alone does not provide sufficient information to estimate the actual quantity consumed or its nutritional composition. This research proposes an enhanced deep learning-based framework for multi-food detection and portion-aware nutritional estimation. The proposed system integrates object detection, image segmentation, food classification, portion-size estimation, and nutritional information retrieval into a unified application. A deep learning-based object detection model is used to identify multiple food items within a single meal image. Individual food regions are subsequently processed to estimate their portions using calibrated image geometry and, where available, depth information. The estimated quantities are combined with nutritional information to calculate the calories, carbohydrates, proteins, and fats associated with each food item. The proposed framework aims to overcome the limitations of conventional single-food classification systems by providing item-wise and meal-level nutritional analysis. Its performance will be evaluated through detection accuracy, segmentation quality, portion-estimation error, and nutritional estimation error
Authors
- Vikram Popat Deokate
- Srushti Deepak Patil Jaydip Jitendra Patil
Publication Details
- Journal
- International Journal of Advanced Research in Science Communication and Technology
- Published
- 2026-11-19
- DOI
- https://doi.org/10.48175/ijarsct-40869
- Primary Topic
- Advanced Neural Network Applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00