RidgeRun is thrilled to announce its new partnership with Coral from Google. Coral is a complete toolkit that allows you to build solutions with on-device inference capabilities that are efficient, private, fast and local. Combined with RidgeRun's expertise in Embedded Linux, Artificial Intelligence and Computer Vision, making smart products is faster and easier than ever before.
We have added support for Coral to GStreamer via GstInference. This allows you to perform seamless predictions on your system's camera, network stream or even recording, in the most efficient way.
Coral support through GstInference allows you to build complex and scalable solutions with lower time-to-market. Not only will you be able to get a working prototype fast, but you will be able to extend it easily with new features. GStreamer's flexibility and modularity will allow you to use the different hardware units in the board making the most out of the platform.
We've created a couple of examples that may be used as reference designs. The first one is a simple media server capable of capturing from either a webcam or an RTSP network stream, and showing these images to a connected display. Then you'll have the capability to start/stop a video recording and network streaming. The fact that this demo is written in a few lines of bash, shows how simple it is to create complex designs for the Coral using GStreamer.
The second example is an intelligent security camera. The application will capture from a camera and show the video in a display. When an interesting object is detected a video recording will be automatically triggered and stopped after the object has left the scene. The list of target objects, as well as other parameters, are configurable. This design, on the other hand, was written in Python and shows how to connect to the prediction signals from the Coral/GstInference.
Make sure you check:
Our developers wiki, for further details on Coral support.
The example pipelines page, as a quick starting point.
The set of examples (hosted on GitHub) for fully functional reference designs!
Technical questions? : support@ridgerun.com
Want to know more about the project? : contactus@ridgerun.com