cnstream/README.md

Cambricon CNStream

CNStream is a streaming framework with plug-ins. It is used to connect other modules, includes basic functionalities, libraries, and essential elements.

CNStream provides the following built-in modules:

• source: Support RTSP, video file, images and elementary stream in memory （H.264, H.265, and JPEG decoding）.
• inference: MLU-based inference accelerator for detection and classification.
• inference2: Based on infer server to run inference, preprocessing and postprocessing.
• osd (On-screen display): Module for highlighting objects and text overlay.
• encode: Encode videos or images.
• display: Display the video on screen.
• tracker: Multi-object tracking.
• rtsp_sink：Push RTSP stream to internet.

Getting started

To start using CNStream, please refer to the chapter of quick start in the document of Cambricon-CNStream-User-Guild-CN.pdf.

Samples

Classification	Object Detection

Object Tracking	License plate recognization

Body Pose	Vehicle Detection

Best Practices

How to build a classic classification or detection application based on CNStream?

You should find a sample from samples/simple_run_pipeline/simple_run_pipeline.cpp that helps developers easily understand how to develop a classic classification or detection application based on CNStream pipeline.

This sample supports typical classification and detection neural networks like vgg resnet ssd fasterrcnn yolo-vx and so on.

This sample supports images or video file as input.

How to change the input video file?

Modify the files.list_video file, which is under the samples directory, to replace the video path. Each line represents one stream. It is recommended to use an absolute path or use a relative path relative to the executor path.

Documentation

Cambricon Forum Docs

Check out the Examples page for tutorials on how to use CNStream. Concepts page for basic definitions.

Community forum

Discuss - General community discussion around CNStream.