VQNetVision is an open-source Python library for Quantum Machine Learning (QML) developed by Origin Quantum, engineered for building, training, and deploying hybrid quantum-classical image models on quantum simulators.
The framework supports fully customizable components—including parameter initialization, quantum data encoders, variational quantum ansatzes, quantum measurements, and classical optimizers—enabling seamless secondary development.
As a lightweight, flexible, and high-performance Quantum Computer Vision (QCV) platform, VQNetVision accelerates algorithm development and quantum image processing research.
It empowers AI researchers, computer scientists, and quantum developers to effortlessly benchmark, simulate, and demonstrate complex computer vision (CV) tasks, particularly quantum image classification and quantum semantic segmentation.
Designed to lower the entry barrier to quantum artificial intelligence (Quantum AI), VQNetVision significantly simplifies the integration of Variational Quantum Circuits (VQCs) into computational vision workflows.
from vqnetvision_pro import QLinear
import torch
ql=QLinear(
in_features=16,
out_features=8,
n_layers=2,
initializer='hadamard_initializer',
encoder='angle_encoder',
ansatz='strong_entangle',
measurement='measurement_expectation',
encoder_kwargs={'gate_sequence': 'rxry'},
measurement_kwargs={'observable': 'Z'},
)
x=torch.randn(4, 16)
y=ql(x)
print('QLinear:', y.shape)
Significantly reduces the number of trainable parameters, mitigates overfitting in neural network training, and improves the generalization performance of the model.
It supports efficient development in a Python environment, features convenient deployment, low system resource usage.
Isolate the networked environment, ensure data circulates internally, and prevent data leakage from affecting data security.
Comes with professional guides and tutorials, enabling even beginners to get started with ease.
Multiple quantum machine learning visual algorithms are provided, supporting function adaptation and customized development for different scenarios.
Greatly boosts the operational performance of visual tasks like image classification, object detection, and image segmentation. It is installed and used via Python wheel packages, which is simple and easy to use, and application scenario examples are provided.
It quickly processes imaging data of structures such as different lobes, bronchi, and blood vessels in the lungs, efficiently localizes lesion areas, and provides a more comprehensive assessment of lung health.
It accurately identifies brain tissues and lesion areas such as tumors, and automatically detects and classifies different types of brain lesions through quantum machine learning models, improving diagnostic accuracy.
It quickly identifies target vehicles in complex traffic backgrounds, accurately judges vehicle violations, and can simultaneously detect target vehicles in terms of safety aspects such as component damage and loss.
It automatically identifies abnormal behaviors of people in surveillance videos, segments items carried by people, judges risks, issues alarms, and improves the effectiveness of surveillance.