Quantum K-Nearest Neighbors for Object Recognition

Object recognition research is essential to simulate human vision capabilities on computers or robots. As time goes by, this research is getting more sophisticated, but it encounters challenges in the form of 3V: (volume) large volume of data; (variety) large variety of data; (velocity); and the need for fast data processing. That matter has led scientists to start looking for solutions to these problems. On the other hand, the development of quantum computing has opened up new opportunities in Quantum Machine Learning (QML), which combines the power of quantum computing with machine learning techniques. One of the exciting algorithms in QML is Quantum k-Nearest Neighbors (QKNN), which can be used in image-based object recognition. However, the use of QKNN in image-based object recognition is still limited and needs to be developed further. This research aims to apply and analyze the quantum computing-based QKNN algorithm in image-based object recognition. The steps include representing the image as quantum states, calculating the distance between two quantum states using the fidelity method, and determining the label using a majority vote based on the closest distance. In this study, the test of QKNN algorithm used 84 synthetic image data sets with a ratio of 64:20. The experimental results on the 2-class variety, the QKNN succeeded on average 0.80, show that the QKNN algorithm can recognize objects with an accuracy rate of 0.65 on the 4-class data set. Based on these results, there is a need for further study in terms of data fidelity and data preprocessing techniques to improve QKNN’s performance.