Quantum Convolutional Neural Network Performance for Efficient Image Analysis and Prediction

Abstract

To study Quantum Convolutional Neural Network (QCNN) performance for Image classification and analysis.

Objective: The core objective is to harness the computational capabilities of quantum systems to expedite traditional machine learning tasks

Methods: Quantum computation is based on quantum mechanics. It uses special properties such as Superposition, Entanglement, and Unitary Transformation which are not there in classical binary systems.

Results: The QCNN model was tested using plant Leaf dataset. The model demonstrated prediction accuracy of 95.45%  with  precision and recall scores of 0.9555 and 0.9545 respectively

Conclusions: The Plant leaf image dataset had been used to quantitatively test the classification performance. The QCNN has presented 95.55% and CNN has given 80.9% of test accuracy with 10 epochs of training on a moderate size dataset, which is a considerable achievement on resource utilization efficiency. In this regard, QCNN outperformed the classical CNN by providing 97.7% of prediction accuracy, in contrast to 67% of prediction accuracy from Classical-CNN as inferred from the confusion matrices.