WiMi Unveils Resource-Efficient QRAM-Based Quantum Convolutional Neural Network to Accelerate Large-Scale Image Classification

WiMi Hologram Cloud Inc. (NASDAQ: WiMi) (“WiMi” or the “Company”), a leading global Hologram Augmented Reality (“AR”) Technology provider, today announced the release of its latest independently developed achievement: a quantum convolutional neural network (QCNN) that classifies images efficiently using quantum random access memory (QRAM). The Company designed this technology for typical image classification tasks that involve large-scale input data and multiple output channels. In doing so, WiMi directly tackles the key bottlenecks that have held back existing quantum convolutional neural networks, namely the number of qubits they require, their circuit depth, and their data loading efficiency, and it charts a feasible path toward applying quantum computing in real machine learning scenarios.

Why Traditional CNNs Hit a Wall

Classical convolutional neural networks succeed on classical computers because they rely on highly parallel matrix operations and massive storage resources. However, as image resolution climbs and the number of feature channels keeps growing, the resource demands of CNNs during both training and inference rise exponentially. Consequently, this trend not only strains hardware but also narrows where these models can realistically operate, particularly in edge computing, low-power devices, and high-real-time settings. To address this, researchers proposed quantum convolutional neural networks, which aim to compress the required computational resources dramatically while preserving model expressiveness through quantum superposition and quantum parallelism.

The Engineering Challenges Facing QCNNs

Even so, existing QCNN models still struggle at the engineering level. On one hand, today’s quantum devices offer only a limited number of qubits and carry high noise levels, which makes handling large-scale input data directly difficult. On the other hand, many quantum neural network schemes rely on amplitude encoding or angle encoding during the data encoding stage, and therefore they must load each input sample individually. Because that process itself consumes a great deal of circuit depth, it ends up offsetting the very advantages that quantum computing promises. As a result, the ability to process large-scale image data in a single, efficient pass under limited quantum resources has become the core problem constraining the practical use of QCNNs.

How WiMi Rethinks Data Access with QRAM

To solve these issues, WiMi re-examined the data access problem in quantum neural networks at the architectural level and introduced quantum random access memory (QRAM) as a key technology within the overall QCNN design. Fundamentally, QRAM uses quantum superposition states to access multiple memory addresses at once, which lets the system index and invoke massive amounts of classical data using only a logarithmic number of qubits. This characteristic makes QRAM a natural fit for the large-scale feature representation that deep learning demands.

In this scheme, WiMi does not treat QRAM as a simple data loading tool. Instead, the Company embeds it deeply into the feature extraction and channel mapping process of the QCNN, thereby forming an entirely new model structure. First, during classical preprocessing, the system maps the input image into a data structure suited to quantum storage. Next, QRAM constructs the corresponding quantum state representation. Unlike traditional pixel-by-pixel or block-by-block loading, this approach allows multiple spatial locations and feature dimensions to exist simultaneously within the quantum state, and therefore it enables parallel access to large-scale input data in a single quantum operation.

Redefining Quantum Convolution

At the convolution computation level, WiMi’s model redefines how quantum convolution kernels work. Traditional QCNNs typically realize convolution operations through local quantum gate combinations, which sharply increase circuit depth as the input scale grows. By contrast, this technology leverages the parallel addressing that QRAM provides to match convolution kernel parameters with input features in the form of quantum states, and it thereby transforms the convolution process into a series of controlled quantum operations. As a result, this design substantially weakens the link between convolution-layer depth growth and input size, which allows the model to keep a relatively shallow quantum circuit even when it processes high-resolution images or multi-channel features.

Expanding Output Channels Without Exhausting Qubits

When it comes to output channel expansion, this technology again reflects a philosophy of resource efficiency. In classical CNNs, adding output channels usually drives linear or even super-linear growth in computation and storage demands. Meanwhile, in quantum architectures, blindly expanding channels would quickly exhaust the available qubit resources. To avoid this, WiMi’s QRAM-based channel mapping mechanism introduces auxiliary index registers into the quantum state, which lets multiple output channels coexist in superposition within the same quantum circuit. The system then reads them out selectively during measurement to complete classification decisions. Consequently, this mechanism sharply reduces the direct quantum resource consumption caused by channel-count growth and enables the model to take on more complex classification tasks.

A Hybrid Quantum-Classical Approach

Throughout training and inference, this QCNN model adopts a typical hybrid quantum-classical architecture. Specifically, the classical side handles parameter updates and loss function evaluation, while quantum circuits carry out the core feature extraction and mapping. This design carefully accounts for the current stage of quantum hardware, and it therefore avoids depending on large-scale fault-tolerant quantum computers while still maximizing quantum computing’s advantages in high-dimensional feature spaces. Through repeated iterative optimization, the model ultimately strikes a strong balance among qubit count, circuit depth, and runtime, all while maintaining classification accuracy.

Validating the Results

During experimental validation, WiMi systematically evaluated the model across multiple image classification tasks of different scales. Notably, the results show that even as input data scale rose significantly and output channel counts expanded continuously, the QRAM-based quantum convolutional neural network outperformed comparable QCNN schemes in both resource consumption and circuit depth while maintaining competitive classification performance. Ultimately, these findings confirm that the technology can handle large-scale machine learning tasks within resource-constrained quantum environments.

Looking Ahead

WiMi’s QRAM-based quantum convolutional neural network for efficient image classification does more than expand the theoretical design space of quantum machine learning models; it also delivers a practically feasible solution for large-scale data processing at the engineering level. Moreover, as quantum hardware capabilities continue to improve, the QRAM-based QCNN architecture should further amplify its parallel computing advantages. In turn, it promises new solutions for complex visual tasks and lays a solid foundation for the real-world deployment of future quantum intelligent systems.

About WiMi Hologram Cloud

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.