# WiMi Hologram Cloud Inc. Unveils H-QNN Technology for Efficient Binary MNIST Image Classification - Link: https://www.thailand-business-news.com/pr-news/wimi-hologram-cloud-inc-unveils-h-qnn-technology-for-efficient-binary-mnist-image-classification - Published: 2026-08-13T21:30:00+07:00 - Author: PR Newswire BEIJING, Aug. 13, 2026 /PRNewswire/ — WiMi Hologram Cloud Inc. (NASDAQ: WIMI) (" WIMI" or the "Company"), a globally leading technology provider, announces a major breakthrough in releasing Hybrid Quantum Neural Network (H-QNN), an innovative hybrid quantum-classical neural network technology tailored for the image recognition sector. The technology creatively integrates parameterised quantum circuits with classical neural network architectures and has been successfully deployed for binary image classification tasks on the MNIST dataset, delivering outstanding performance in classification accuracy, feature representation capability, and model training efficiency. Image classification stands as one of the most fundamental and critical tasks within the field of computer vision. Spanning handwritten digit recognition, facial recognition, industrial defect detection, and autonomous driving perception systems, image classification technology forms the core foundation of nearly all modern AI visual systems. Conventional deep learning models primarily rely on Convolutional Neural Networks (CNNs) to execute feature extraction. CNNs extract edge, texture, contour, and semantic information step-by-step via sliding convolution kernels across images, then leverage fully connected networks to render classification decisions. Nevertheless, as data dimensions expand and image features grow increasingly intricate, traditional neural networks have gradually revealed multiple inherent limitations. First, data distributions within high-dimensional feature spaces often feature complex nonlinear structures, requiring classical networks to incorporate massive quantities of parameters to construct sufficiently sophisticated decision boundaries. Second, deep network training is prone to issues such as vanishing gradients, local optima, and overfitting. In addition, training large-scale models demands enormous computational resources and energy consumption. Meanwhile, advances in quantum computing have opened new avenues to address the aforementioned challenges. Its core design philosophy centers on fully harnessing the high-dimensional feature mapping capacity of quantum circuits during image classification: complex pattern recognition tasks are delegated to quantum layers, while parameter optimisation and final classification decisions are handled by mature, stable classical neural networks. This architecture design effectively circumvents the constraints imposed by the limited scale of current quantum hardware, while maximising the inherent advantages of quantum computing in feature representation. From an overall architectural perspective, WIMI’s H-QNN establishes a complete end- to-end data processing pipeline that enables deep integration between quantum computing and classical computing. To render classical images processable by quantum computers, the conversion of classical data into quantum-compatible data must be resolved first. Each image in the MNIST dataset consists of a 28×28-pixel grid with 784 grayscale values in total. Directly loading all pixel data into a quantum system would incur prohibitive quantum resource overhead. Accordingly, H-QNN first employs a classical preprocessing module to conduct dimensionality reduction and normalisation on input images. Standardised feature vectors are subsequently converted into data formats compatible with quantum state representation before entering the quantum encoding phase. Within this phase, classical features are mapped to the amplitudes and rotation angles of quantum bits. Each qubit is initialised to the ground state 0, and feature encoding is implemented through rotation gates. Quantum rotation operations translate image features into quantum state parameters; following this transformation, pixel information from the original image is embedded within the quantum state space, laying the groundwork for subsequent quantum feature learning. One of the most pivotal innovations of WIMI’s H-QNN lies in its parameterised quantum feature learning module. Where traditional CNNs use convolution kernels to learn image features, this responsibility is undertaken by parameterised quantum circuits in H-QNN. The quantum layer is composed of stacked rotation gates and entanglement gates: rotation gates execute local feature transformations, whereas entanglement gates establish correlation relationships between distinct qubits. Quantum states continuously evolve throughout this process. Driven by quantum entanglement mechanisms, intricate correlational structures emerge across multiple qubits, a correlation capacity far surpassing the linear connection schemes adopted by traditional neural networks. When certain patterns within images carry complex spatial relationships, quantum entanglement naturally captures these high-order features. This mechanism is particularly well-suited to processing nonlinear distributions embedded within high-dimensional data. Complex feature representations that may require hundreds or even thousands of neurons in classical networks can be expressed with far fewer parameters within the quantum feature space. The essence of classification lies in identifying decision boundaries, which traditional neural networks construct through multi-layer nonlinear transformations. By contrast, WIMI’s H-QNN leverages the exponential dimensional advantage of quantum state spaces to deliver a more flexible classification mechanism. Within quantum state space, samples belonging to different categories are mapped to distinct spatial regions. Following multi-layer quantum transformations, data points that prove difficult to separate in classical space become far more distinguishable. Quantum feature mapping effectively amplifies inter-class distances, lowering classification difficulty; such quantum-enhanced feature spaces can substantially boost model recognition performance. Upon completing quantum feature learning, the quantum system transmits extracted information back to the classical neural network via quantum measurement. The quantum measurement layer retrieves quantum state information and converts it into classical numerical values. Specifically, feature vectors are obtained by measuring the expectation values of Pauli operators, which are then fed into the classical classification network. The measurement process functions to compress and project high-dimensional features learned within quantum space back into classical space. Since complex feature extraction has already been completed by the quantum layer, the subsequent classification network can maintain a compact structure, allowing the overall model to sustain high precision while cutting down computational overhead. Moving forward, WIMI intends to extend this model to more complex datasets. The company will also explore deeper quantum network architectures, adaptive quantum feature extraction mechanisms, and large-scale quantum entanglement learning frameworks. As quantum hardware performance sees sustained improvements, hybrid quantum-classical neural networks are poised to transition from laboratory research to industrial deployment, delivering transformative value across intelligent manufacturing, autonomous driving, remote sensing recognition, biomedical treatment, and financial analytics. The launch of H-QNN not only demonstrates the immense potential of quantum computing to empower artificial intelligence but also charts a new development trajectory for next-generation intelligent computing architectures, marking a vital milestone as quantum-enhanced machine learning advances from theoretical research toward practical real-world applications. **About WiMi Hologram Cloud Inc.** WiMi Hologram Cloud Inc. (NASDAQ: WIMI) focuses on holographic cloud services, primarily concentrating on professional fields such as in-vehicle AR holographic HUD, 3D holographic pulse LiDAR, head-mounted light field holographic devices, holographic semiconductors, holographic cloud software, holographic car navigation, metaverse holographic AR/ VR devices, and metaverse holographic cloud software. It covers multiple aspects of holographic AR technologies, including in-vehicle holographic AR technology, 3D holographic pulse LiDAR technology, holographic vision semiconductor technology, holographic software development, holographic AR virtual advertising technology, holographic AR virtual entertainment technology, holographic ARSDK payment, interactive holographic virtual communication, metaverse holographic AR technology, and metaverse virtual cloud services. WiMi is a comprehensive holographic cloud technology solution provider. For more information, please visit [http://ir.wimiar.com](http://ir.wimiar.com/). **Translation Disclaimer** The original version of this announcement is the officially authorized and only legally binding version. If there are any inconsistencies or differences in meaning between the Chinese translation and the original version, the original version shall prevail. WiMi Hologram Cloud Inc. and related institutions and individuals make no guarantees regarding the translated version and assume no responsibility for any direct or indirect losses caused by translation inaccuracies. --- | This article was produced by Cision PR Newswire, our trusted news partner. The views expressed and the content presented here are solely those of the author and may not fully reflect the opinions of Thailand Business News. | --- **Read the original article :** [WiMi Hologram Cloud Inc. Unveils H-QNN Technology for Efficient Binary MNIST Image Classification ](http://www.prnasia.com/story/archive/5025977_CN25977_0?rand=184833)