Are you looking for a quick and easy way to implement object detection and tracking? RidgeRun and Arducam have worked together to launch a media server and a docker container that uses the IMX708 camera and is capable of performing AI inference on live video in order to detect and track multiple objects available on the video stream that is being received by the Jetson.
Features available on the demo?
We have developed a media server that lets the user interact with the following key features:
Display video with the bounding boxes of the objects detected and their respective tracking. (Objects: Persons, cars, bicycles and traffic signs)
Start video recordings.
Take snapshots. Whether there is a recording running or not.
Stop video recordings.
Record multiple videos.
Get started into the AI world with RidgeRun’s and Arducam’s NVIDIA Jetson Demo!
Demo in real life sample.
Take a look at a demo video on the left. It was recorded using the NVIDIA Jetson Orin NX and the Arducam IMX708 camera along with the docker container and the media server running in order to perform the object detection and tracking. It shows the execution of the demo in a day to day environment, detecting and tracking cars, people, bicycles and traffic signs.
Deep dive
The following diagram shows a segmented view of the media server. The initial section is the video capture pipeline, which is in charge of using the IMX708 from Arducam to capture the video stream. Then the video processing and Deepstream pipelines come into action, the video processing adapts the video stream to a format that Deepstream understands and the Deepstream pipeline is the one that uses multiple elements to detect, track and draw the bounding boxes over the objects on the video stream. The encoding section manages to use the hardware encoders of the NVIDIA Jetson Orin NX to encode the video stream coming from the Deepstream pipeline and send it in a known format to the final section of the media server which contains key functionalities. The final features of the media server are the recording, which can be performed in H264 and H265 encoding and the snapshot which is performed on JPEG encoding.
Demo Setup
On the hardware side, this demo was performed with a NVIDIA Jetson Orin NX and Arducam’s IMX708 cameras, along with the cables required for connecting the hardware. On the software side the demo needs RidgeRun’s IMX708 open source driver, the docker container that has all the dependencies satisfied for executing the demo and the media server which is in charge of executing the pipelines of the section explained above in order to put it all together and perform the object detection and tracking.
If you're interested in buying one of these awesome cameras please check Arducam’s IMX708 product website.
Feel free to check RidgeRun’s developer wiki for instructions on how to install and execute the demo.
What’s Next?
Stay tuned and check out our developer wiki for more information.