Intelligent Deep Learning-based Approach for Non-destructive Food Freshness Diagnosis

Artificial intelligence recognition combines spectroscopy with machine learning models through mobile phones or portable spectroscopic devices that provide dish recognition information from images where a single dish appears, providing useful nutrition-related information.




The mobile application first recognizes the food items inside an overall meal dish in real-time by taking an image. Then, sequentially a conformity testing search queries were based on “food recognition,” “food classification,” “food portion estimation,” “food logging,” and “food image dataset.” This classification is considered the first semantic level (food types detection) and combines with AI-based data analytics, and it extra classified the recipe.


NIR spectroscopy provided the spectral signatures to analyze ingredient composition. The concentrations of protein, fat, moisture, carbohydrates, salt, and fibers can be easily diagnosed with NIR spectroscopy, making it ideal for screening final products, e.g., labeling claims.

Concurrently, a deep convolutional neural network (CNN) architecture is automatically applied to extract fresh features from images.


In conclusion, the proposed CNN-based method has lower complexity with higher accuracy compared to traditional classification methods. Therefore, this method is well-capable for monitoring and classifying freshness as a fast, low-cost, precise, non-destructive, real-time, and automated technique. Furthermore, results showed a classification accuracy could reach more than 95%.


The acquired mobile applications promote healthier lifestyle choices.


Schedule a short zoom session to talk to us now.



17 views0 comments