Universal Food Detection and Nutritional Analysis System using Deep Learning and Explainable AI Models
Food is an essential part of human life, and the nutritional value of our diet is important for maintaining a healthy lifestyle. The traditional method of manually tracking food intake is not only tedious but also error prone. To address this, we propose a universal food detection, classification, segmentation, and 3D volume estimation system based on deep learning models to extract nutritional information from images of food.
The primary aim of this proposal is to develop a universal food detection and analysis system based on deep learning models to accurately classify, segment, and estimate the volume of various types of food, including simple and complex foods. As a result, this system will be able to extract the nutritional information of the intake food from the image and provide explainability based on generative AI to produce accurate reports before and after serving the food. To achieve the objectives mentioned above, we will use deep learning models that be trained on large food databases. The dataset will be provided by the industry partner. The models will be able to recognize different types of food and classify them based on their nutritional content. We will develop novel deep learning model for food detection and segmentation. We will also use object detection algorithms to detect and localize the food in the image. To estimate the volume of the food in the image, we will develop a unique deep learning algorithm based on a combination of depth estimation and image segmentation techniques. The model will be trained using the proposed model on a large dataset of images with known volumes of food to accurately estimate the volume of the food in the image. A volume-based nutrition calculation will be used to determine the food's nutrition intake.
Project Requirements and Goals
- A generative AI model will be used to explain nutrition intake before and after serving.
- Explainable AI models are designed to provide transparency and provide more information to users understand and interpret the results of the nutritional analysis.
Scholarship Type:
A PhD Student (3-3.5 Years)
Host University:
Deakin University
Industry Partner:
AerVision is a technology company developing innovative solutions for enhanced safety and security using state-of-the-art artificial intelligence, computer vision, biometrics, Internet-of-Things, and video analytics. The company was founded in 2013 by two world-leading PhD biometrics and AI scientists, with the focus on helping organisations improve safety, security, and optimising operations. Today, AerVision deploys a number of cutting-edge applications in a number of industries such as aviation, education, correctional facilities, healthcare, government, commercial, mining, and retail.