OpenCV 5.0 Released with Rewritten DNN Engine, Built-In LLM and VLM Support (www.phoronix.com)

OpenCV has officially launched version 5.0, marking a significant advancement in the widely-used library for real-time computer vision and machine learning applications. This release introduces a completely rewritten deep neural network (DNN) engine, which now boasts over 80% support for ONNX models, enhancing interoperability and performance. Notably, OpenCV 5.0 integrates built-in support for large language models (LLM) and vision language models (VLM), broadening the scope of applications it can effectively handle. Additionally, a new hardware abstraction layer and an improved 3D vision toolkit have been introduced, optimizing performance on various platforms including Intel, Arm, Qualcomm, and RISC-V. The implications of this update are substantial for the AI/ML community, as it positions OpenCV as a more competitive tool within the landscape of AI frameworks, particularly against Microsoft’s ONNX Runtime. The introduction of GPU support is also on the horizon, promising even greater speed and efficiency. This release not only enhances the capabilities of OpenCV for developers and researchers but also opens up new possibilities for innovative applications in computer vision and beyond. OpenCV 5.0 is now available for download on GitHub.