Quantum Pre-Processing for Image Recognition
This study investigates a method for image classification using hybrid quantum-classical neural networks. A quantum pre-processing step is introduced to enhance the feature extraction process. A quantum convolutional algorithm is first applied to the images, and a classical neural network is employed to perform classification. Results confirm that preparing input data using quantum methods enables faster convergence and provides better accuracy in the early learning phases. However, it was also noticed that when this approach is paired with a traditional Convolutional Neural Network (CNN), while showing improvements, they aren’t as pronounced, indicating that there might be some overlap in feature extraction capabilities. The research emphasises the importance of quantum pre-processing for image classification, especially when using noisy or distorted data, marking a distinct contribution by evaluating robustness under combined noise types—a relatively unexplored aspect in prior studies on Quanvolutional networks.
Recent Comments