Skip to main content
The Qualcomm IM SDK GStreamer plugins interact with Qualcomm IM SDK multimedia components to build multimedia or AI/ML use cases. The Qualcomm IM SDK hides the complexity of the hardware within the plugin architecture and provides APIs to applications. Using this framework, you can create applications without accessing low-level platform libraries and hardware details, which can vary across platforms. You can select the Qualcomm IM SDK GStreamer pipeline deployment option directly in your project’s Deployments page:

Qualcomm deployment options

Only YOLO-based models are supported using this deployment option.
When selecting this option, you will obtain a .zip folder with a directory structure like the following:
This folder contains example pipelines and applications to run Edge Impulse models using the Qualcomm Intelligent Multimedia SDK GStreamer plugins. These example pipelines have been tested on the Qualcomm Dragonwing™ RB3 Gen 2 Development Kit.

​
Gstreamer pipeline using a camera

Start via:
This renders both the camera feed from the RB3’s built-in camera plus the inference result (bounding boxes) to the screen connected over HDMI.

​
Gstreamer pipeline using a video file

Start via:
This takes an mp4 video file in, and writes a new video file with both the input video stream, plus the inference result (bounding boxes).

​
Prebuilt edge-impulse-object-detection demo app

Run the application via:
This app renders both the camera feed from the RB3’s built-in camera plus the inference result (bounding boxes) to the screen connected over HDMI.

​
Custom application with the IM SDK C++ SDK

We’ve included a demo application based on gst-ai-object-detection in imsdk-demo-app/. To build:
  1. Build the application (will be built using Docker):
  2. Run the application:

​
Additional resources