Deep Neural Networks for the Classification of Pure and Impure Strawberry Purees

Area of Science:

• Food Science
• Machine Learning
• Spectroscopy Analysis

Background:

• Food adulteration is a significant concern, impacting consumer safety and industry integrity.
• Accurate detection of adulterants in food products like purees is crucial.
• Deep neural networks (DNNs) offer promising avenues for complex classification tasks.

Purpose of the Study:

• To comparatively evaluate the effectiveness of different deep neural networks (DNNs) for classifying pure and adulterated strawberry purees.
• To assess the performance of Gated Recurrent Unit (GRU), Long Short Term Memory (LSTM), and Temporal Convolutional Network (TCN) models.
• To identify the most accurate DNN for detecting adulteration in time series spectroscopy data.

Main Methods:

• Utilized the Strawberry dataset, a time series spectroscopy dataset from the UCR time series classification repository.
• Employed three distinct DNN architectures: GRU, LSTM, and TCN.
• Evaluated model performance based on classification accuracy for pure vs. impure puree samples.

Main Results:

• The Temporal Convolutional Network (TCN) demonstrated superior classification accuracy compared to both GRU and LSTM models.
• The TCN achieved a new state-of-the-art classification accuracy on the Strawberry dataset.
• These findings highlight the TCN's effectiveness in identifying subtle patterns indicative of adulteration.

Conclusions:

• The TCN is a highly effective deep learning model for the detection of adulteration in strawberry purees.
• The study indicates significant potential for TCN application in the broader field of fruit puree adulteration detection.
• Advanced deep learning models like TCN can enhance food safety and quality control measures.