Research pick: Squirrelling away nutritional information - "Food recognition using enhanced squirrel search optimisation algorithm and convolutional neural network"

Researchers in India have developed a search algorithm based on the strategy used by squirrels to find their cached nuts to automate food identification. Details are described in the International Journal of Data Analysis Techniques and Strategies. The algorithm could have applications in the food industry, hospitality and even in a dietary healthcare environment.

Megha Chopra and Archana Purwar of the Department of Computer Science Engineering and Information Technology at Jaypee Institute of Information Technology in Noida, explain how a system to identify and assess food and tie the item to known nutritional profiles for that food could have many uses in a range of areas.

The best recipe for food recognition begins with the classification of food items from images. This classification process is initiated by segmenting the food images, a standard step in image analysis. Conventionally, thresholding is used in segmentation, however, Chorpa and Purwar have taken a novel approach. They use a Squirrel Search Algorithm (SSA) to optimize multi-level thresholding. This SSA-based optimization is designed to improve the accuracy of food image recognition. A Convolutional Neural Network (CNN), a powerful artificial intelligence tool, then classifies the food in the images.

The team reports significant improvements on earlier approaches to image segmentation and food item classification, achieving an accuracy rate of well over 80 per cent in tests; specifically up to 83.1%.

Accurate food recognition is a pivotal component of automatic dietary assessment. With improved segmentation and classification, tools could be developed for companies, healthcare providers, and individuals to monitor “calorie” intake, and nutritional value, and make better informed dietary choices.

The same tools might also embed recognition and flagging systems for allergenic or problematic foods in a dish and so help protect individuals from inadvertently eating something that might lead to an allergic response or to which they have an intolerance.

In summary, this research presents a novel approach to automated food recognition, offering not only a technical advancement on earlier approaches but also promising possibilities for improved dietary assessment and personalized dietary management.

Chopra, M. and Purwar, A. (2023) ‘Food recognition using enhanced squirrel search optimisation algorithm and convolutional neural network’, Int. J. Data Analysis Techniques and Strategies, Vol. 15, No. 3, pp.238–254.