QUANTIC (QUantum ANalysis and Technology for Image Classification)
Artificial Intelligence technologies for digital image processing are becoming increasingly pervasive across all industrial sectors. Among these technologies, classification plays a fundamental role in the accurate interpretation of visual data. This project explores the integration of these AI methodologies with Quantum Computing (QC), aiming to push the boundaries of what’s possible in image classification through Quantum Machine Learning (QML). With QC entering the era of practical utility, comparing classical and quantum algorithms is critical to assess the true capabilities of quantum approaches in real-world scenarios. Our primary goal is to develop and implement QML models, specifically Quantum Convolutional Neural Networks (QCNNs), for image classification tasks. We will benchmark these models against their classical counterparts evaluating key factors such as accuracy, computational resources, and scalability. As a real-world case study, we focus on the classification of hazelnut leaves and fruits to detect varietal differences: a task complex enough to challenge state-of-the-art classical models. The project aims to:
- Identify the resource thresholds where quantum hardware limitations begin to hinder the reliability of quantum algorithms’ executions;
- Understand how computational requirements scale with increasing problem size;
- Establish robust benchmarks for evaluating QML performance in image classification.
This initiative is a collaboration between QuantumNet, Netcom’s R&D division and academic partners (University of Salerno), fostering synergy between research and technological advancement. By tracking progress in both quantum hardware and software development, the project contributes to a deeper understanding of QML’s industrial potential and, at the same time, faces one of the most challenging problems in agri-food industry.
Key Innovations:
- New QCNN Architectures
We will design an advanced QCNN model tailored for classifying hazelnut varieties, merging quantum principles with deep learning techniques to enhance classification performance. - Comprehensive Benchmarking
A rigorous benchmarking framework will be developed to objectively compare QML and classical ML models using metrics like classification accuracy, efficiency, and scalability. - Quantum Scalability Analysis
We will identify critical limits where quantum hardware constraints impact solution quality, offering valuable insights for future hardware and software design. - Pioneering Agricultural Application
This represents one of the first uses of QML in agriculture, highlighting its potential for revolutionizing plant health monitoring and crop quality assessment.
By addressing both theoretical and practical aspects, the project contributes to the ongoing discourse on quantum advantage and helps bridge the gap between quantum research and real-world applications.