Quantum Neural Networks in Practice: A Comparative Study with Classical Models from Standard Data Sets to Industrial Images

Image classification tasks are among the most prominent examples that can be reliably solved by classical machine learning models. In this study, we compare the performance of randomized classical and quantum neural networks as well as classical and quantum-classical hybrid convolutional neural networks for the task of binary image classification. To this end, we employ various data sets of increasing complexity - (i) an artificial hypercube dataset, (ii) MNIST handwritten digits, and (iii) real-world industrial images from laser cutting machines. We analyze the performance of the employed quantum models with respect to correlations between classification accuracy and various hyperparameters. For the random quantum neural networks, we additionally compare their performance with some known literature models and how top-performing models from one data set perform on the others. In general, we observe fairly similar performances of classical and quantum or hybrid models. Our study provides an industry perspective on the prospects of quantum machine learning for practical image classification tasks.

@article{basilewitsch2025_2411.19276,
  title={ Quantum Neural Networks in Practice: A Comparative Study with Classical Models from Standard Data Sets to Industrial Images },
  author={ Daniel Basilewitsch and João F. Bravo and Christian Tutschku and Frederick Struckmeier },
  journal={arXiv preprint arXiv:2411.19276},
  year={ 2025 }
}