The AM6x Linux SDK provides GStreamer (GST) framework, which can be used in user space to integrate the image processing components for various applications. The Hardware Accelerators (HWA) on the SoC, such as the Vision Pre-processing Accelerator (VPAC) or ISP, video encoder and decoder, and deep learning compute engine, are accessed through GST plugins. The VPAC (ISP) has multiple blocks, including Vision Imaging Sub-System (VISS), Lens Distortion Correction (LDC), and Multiscalar (MSC), each corresponding to a GST plugin.

Figure 3-2 shows the block diagram of a typical image pipeline from the camera to encoding or deep learning applications on AM62A. For more details about the end-to-end data flow, refer to the EdgeAI SDK documentation.

Figure 3-2 A Typical AM62A Image Pipeline Using GStreamer

For AM62P, the image pipeline is simpler because there is no ISP on AM62P.

Figure 3-3 A Typical AM62P Image Pipeline Using GStreamer

With a video node created for each of the cameras, the GStreamer-based image pipeline allows the processing of multiple camera inputs simultaneously. The multiple cameras can be the same or different cameras. For AM62A, the ISP is reconfigured frame-by-frame to process images from each connected camera. A reference design using GStreamer for multi-camera applications is given in the next chapter.