Scholarlink

<p>Recognition of food images automatically plays a significant role in health-centric applications, including diet assessment and allergy detection. Correct recognition of food images can assist people with allergic reactions to avoid restricted nutrients, as well as enable smart nutrition control systems. In this work, we propose a deep learning-driven model for visual recognition of food and allergen identification using different convolutional neural network (CNN) structures such as MobileNet, DenseNet121, VGG19, ResNet50, and EfficientNetB0. The models are comparably assessed in terms of different standard performance measures. The performance metrics of accuracy 99.41%, precision 99.42%, recall 99.41%, and F1-score 99.42% were obtained by EfficientNetB0 as the best-performing model among all tested models. Due to its effectiveness and high reliability, EfficientNetB0 was the chosen one as the final model implemented in a Streamlit-based application for online food recognition and allergen classification, proving the practical applicability of the presented system in a real-world scenario.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automated food visual recognition and allergen identification using deep learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月28日

查看全文
  • Rajesh Natarajan,
  • Sujatha Krishna,
  • Pradeepa Ganesan,
  • Saleem Raja Abdul Samad,
  • Syed Ibrahim Hussain,
  • Myra Patalay

摘要

Recognition of food images automatically plays a significant role in health-centric applications, including diet assessment and allergy detection. Correct recognition of food images can assist people with allergic reactions to avoid restricted nutrients, as well as enable smart nutrition control systems. In this work, we propose a deep learning-driven model for visual recognition of food and allergen identification using different convolutional neural network (CNN) structures such as MobileNet, DenseNet121, VGG19, ResNet50, and EfficientNetB0. The models are comparably assessed in terms of different standard performance measures. The performance metrics of accuracy 99.41%, precision 99.42%, recall 99.41%, and F1-score 99.42% were obtained by EfficientNetB0 as the best-performing model among all tested models. Due to its effectiveness and high reliability, EfficientNetB0 was the chosen one as the final model implemented in a Streamlit-based application for online food recognition and allergen classification, proving the practical applicability of the presented system in a real-world scenario.