Introduction

This example demonstrates using the VAAL plugins for GStreamer to run a detection model and capturing bounding box results from Python.  While this example is provided as Python code, we also provide C and command-line examples in the following articles.

• Detection from C
• Detection from Command-Line

Using GStreamer from Python is done with the PyGObject library which makes use of the GObject Introspection mechanism to enable cross-language interoperability.  The VAAL library can automatically handle various types of objects such as those used by GStreamer to carry camera (or video file) frames.

Requirements

• NXP i.MX 8M Plus EVK
• OV5640 Camera connected by MIPI
• NXP Yocto BSP 5.15.32-2.0.0
• VisionPack 1.2.2 or newer.

Usage

The application at a minimum needs a model to run.  You can use any detection model trained in eIQ Portal or ModelPack, we provide a person detection model as an attachment to this article to evaluate.

gst-detect.py modelpack-people-320x320.rtm

The application will run and provide timing and detection results on the output console.

Drawing Results

*COMING SOON*

Downloads

The gst-detect.py example is provided in the VisionPack bundle and through the vaal-samples package.

• VisionPack Downloads