INFORMAZIONI SU
BIANCO Rachele
 |
Supervisore: Prof.ssa Parpinel
|
| Background: inadequate diet is considered one of the principal risk factors in non-communicable diseases. However, because of several limitations in data collection, future challenges are focused to design new tools for diet assessment using innovative approaches. Machine learning (ML) is a subfield of artificial intelligence that enables computers to learn without being directly programmed, and its application has been recently considered in nutritional epidemiology. In detail, deep learning can be used to automatically classify foods from pictures. These techniques may improve precision and validity in diet records, overcoming self-report biases. Methods: a pilot dataset including Italian recipes and detailed related information will be realized. An algorithm based on a specific procedure including Convolutional Neural Network models will be developed. Finally, the training of the ML algorithm on the Nutrition5k food dishes dataset and the development and validation of an innovative image-based dietary assessment tool will be carried out. Conclusion: despite of its excellent performance in several fields, deep learning has never been applied to develop dietary assessment tools in Italy. Expected results include ML algorithms to assess weights from food images to support diet monitoring and open Italian atlas of food dishes with the related nutritional composition of recipes. | |