VideoPipe — конвейер компьютерного зрения для видеоаналитики

★ 2.9K

VideoPipe is a C++ framework for video analysis and structuring, built as a pipeline of independent nodes you combine to fit your task. Reach for it when you need to assemble your own video-analytics application: object detection and recognition on a stream (including all YOLO versions), face recognition, behavior analysis, image search, and traffic/security incident analysis. In spirit it resembles NVIDIA DeepStream and Huawei mxVision, but it is easier to use and more portable, with minimal dependencies. Its focus is a framework for building computer-vision pipelines over video streams, not a finished surveillance app or a standalone model: you connect nodes (capture, detection, tracking, output) for a specific scenario. It is useful to developers of video-analytics systems who need a controllable pipeline rather than a boxed product.