Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-Modal Zero-Shot Prediction of Color Trajectories in Food Drying

Shichen Li, Ahmadreza Eslaminia, Chenhui Shao

TL;DR

<3-5 sentence high-level summary>Addressing the challenges of predicting color-change trajectories in food drying under unseen processing conditions, the paper introduces a multi-modal, zero-shot framework that represents color evolution as a weighted sum of predefined basis functions with learned coefficients. The approach integrates DCT-based denoising, initial-state image features via multi-modal fusion, and similarity-based training data selection to enhance generalization, outperforming a baseline moment-by-moment predictor. Case studies on cookie and apple drying demonstrate strong zero-shot generalization and substantial RMSE improvements, highlighting robust performance and practical potential for in-line quality monitoring. The work advances trajectory-aware quality control in industrial drying by leveraging high-dimensional color data, process parameters, and early visual cues to generalize across varied conditions.

Abstract

Food drying is widely used to reduce moisture content, ensure safety, and extend shelf life. Color evolution of food samples is an important indicator of product quality in food drying. Although existing studies have examined color changes under different drying conditions, current approaches primarily rely on low-dimensional color features and cannot fully capture the complex, dynamic color trajectories of food samples. Moreover, existing modeling approaches lack the ability to generalize to unseen process conditions. To address these limitations, we develop a novel multi-modal color-trajectory prediction method that integrates high-dimensional temporal color information with drying process parameters to enable accurate and data-efficient color trajectory prediction. Under unseen drying conditions, the model attains RMSEs of 2.12 for cookie drying and 1.29 for apple drying, reducing errors by over 90% compared with baseline models. These experimental results demonstrate the model's superior accuracy, robustness, and broad applicability.

Multi-Modal Zero-Shot Prediction of Color Trajectories in Food Drying

TL;DR

<3-5 sentence high-level summary>Addressing the challenges of predicting color-change trajectories in food drying under unseen processing conditions, the paper introduces a multi-modal, zero-shot framework that represents color evolution as a weighted sum of predefined basis functions with learned coefficients. The approach integrates DCT-based denoising, initial-state image features via multi-modal fusion, and similarity-based training data selection to enhance generalization, outperforming a baseline moment-by-moment predictor. Case studies on cookie and apple drying demonstrate strong zero-shot generalization and substantial RMSE improvements, highlighting robust performance and practical potential for in-line quality monitoring. The work advances trajectory-aware quality control in industrial drying by leveraging high-dimensional color data, process parameters, and early visual cues to generalize across varied conditions.

Abstract

Food drying is widely used to reduce moisture content, ensure safety, and extend shelf life. Color evolution of food samples is an important indicator of product quality in food drying. Although existing studies have examined color changes under different drying conditions, current approaches primarily rely on low-dimensional color features and cannot fully capture the complex, dynamic color trajectories of food samples. Moreover, existing modeling approaches lack the ability to generalize to unseen process conditions. To address these limitations, we develop a novel multi-modal color-trajectory prediction method that integrates high-dimensional temporal color information with drying process parameters to enable accurate and data-efficient color trajectory prediction. Under unseen drying conditions, the model attains RMSEs of 2.12 for cookie drying and 1.29 for apple drying, reducing errors by over 90% compared with baseline models. These experimental results demonstrate the model's superior accuracy, robustness, and broad applicability.

Paper Structure

This paper contains 21 sections, 6 equations, 14 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Zero-Shot Learning
  4. Trajectory Prediction
  5. Data-Driven Modeling
  6. Methodology
  7. Model Architecture
  8. Multi-Modal Fusion of Initial Stage Data
  9. Similarity-Informed Selection of Training Datasets
  10. Baseline Model
  11. Case Studies
  12. Color Trajectory Prediction in Cookie Drying
  13. Experiments
  14. Data Preparation
  15. Results
  16. ...and 6 more sections

Figures (14)

  • Figure 1: Zero-shot learning-based data splitting strategy.
  • Figure 2: Overall model architecture of data-driven trajectory prediction.
  • Figure 3: Comparison between (a) multi-modal fusion encoder and (b) tabular-only encoder.
  • Figure 4: Similarity-informed selection of training datasets.
  • Figure 5: Experimental setup for cookie drying.
  • ...and 9 more figures