VNCV — Vietnamese Computer Vision is an OCR library built with the goal of optimizing the recognition of Vietnamese text from images. The library allows for quick and convenient extraction of textual content from images while automatically handling necessary steps such as model loading, environment configuration, and inference. As a result, users can integrate OCR capabilities into their systems with just a few lines of code, without the need for complex installation steps.

VNCV is designed to operate efficiently on both CPU and GPU, making it particularly suitable for environments with limited resources. Notably, the library has completely migrated to the ONNX Runtime Engine, eliminating reliance on the relatively heavy PyTorch. This change significantly reduces the installation size of the system, while also noticeably improving processing speed on the CPU. This makes VNCV a lightweight, fast, and easy-to-deploy OCR solution for various types of systems.

When using VNCV, users only need to provide the path to the image that needs recognition. The library automatically downloads the necessary models from the release source if they are not already present on the machine, then proceeds to detect regions containing text in the image and recognize the content of each region. The results can be returned in the form of a text list or as detailed JSON, including the text content, confidence level, and bounding box coordinates. Thanks to the clear data structure, the OCR results can be easily utilized in data analysis systems, document processing, or building service APIs.

In addition to usage through Python, VNCV also provides a command-line interface (CLI), allowing users to run OCR directly from the terminal without writing code. Additionally, the library supports various advanced options such as saving images with marked text regions using bounding boxes or exporting results in JSON format for integration with other systems.

With a lightweight architecture and high automation capabilities, VNCV is especially suitable for applications such as OCR systems for Vietnamese documents, invoice recognition, administrative papers, citizen identity cards, or real-time image processing services. The library is also suitable for deployment in serverless environments, microservices, edge devices, or API systems that require quick response times while still saving resources.

The project is developed by DevHub Solutions and released as open-source for educational and research purposes. VNCV inherits and leverages the power of the recognition engine from the VietOCR project, while optimizing it to better fit practical deployment in modern systems. Users are free to use, learn from, and integrate the library into personal or commercial projects, provided they retain copyright information and acknowledge the project's source.

Thanks to the combination of high performance, lightweight architecture, and flexible integration capabilities, VNCV aims to become a simple yet powerful OCR tool for applications dealing with Vietnamese text processing in practice.