Navigating Weight Prediction with Diet Diary

TL;DR

This paper introduces the DietDiary dataset and a novel framework for predicting future weight based on daily dietary intake, leveraging historical food data and weight measurements with state-of-the-art time series models.

Contribution

It presents a new dataset and a model-agnostic forecasting framework with a meal representation module and diet-aware loss for weight prediction.

Findings

Our framework outperforms baseline models in weight prediction accuracy.

The UMRL module effectively captures meal representations.

The diet-aware loss improves correlation between food intake and weight changes.

Abstract

Current research in food analysis primarily concentrates on tasks such as food recognition, recipe retrieval and nutrition estimation from a single image. Nevertheless, there is a significant gap in exploring the impact of food intake on physiological indicators (e.g., weight) over time. This paper addresses this gap by introducing the DietDiary dataset, which encompasses daily dietary diaries and corresponding weight measurements of real users. Furthermore, we propose a novel task of weight prediction with a dietary diary that aims to leverage historical food intake and weight to predict future weights. To tackle this task, we propose a model-agnostic time series forecasting framework. Specifically, we introduce a Unified Meal Representation Learning (UMRL) module to extract representations for each meal. Additionally, we design a diet-aware loss function to associate food intake with weight variations. By conducting experiments on the DietDiary dataset with two state-of-the-art time series forecasting models, NLinear and iTransformer, we demonstrate that our proposed framework achieves superior performance compared to the original models. We make our dataset, code, and models publicly available at: https://yxg1005.github.io/weight-prediction/.