Gst-nvsegvisual

The Gst-nvsegvisual plugin visualizes segmentation results. Segmentation is based on image recognition, except that the classifications occur at the pixel level as opposed to the image level as with image recognition. The segmentation output size is generally same as the input size. For more information, see the segmentation training reference at: https://github.com/qubvel/segmentation_models

Inputs and Outputs

• Inputs

  • GStreamer buffer containing NV12/RGBA frame(s)
  • NvDsInferSegmentationMeta containing class numbers, pixel class map, width, height, etc. generated by Gst-nvinfer.
  • gpu-id: selects the GPU ID
  • width, set according the segmentation output size
  • height, set according the segmentation output size

• Output

This plugin allocates different colors for different classes. For example, the industrial model’s output has only one representing defective areas. Thus, defective areas and background have different colors. The semantic model outputs four classes with four different colors: car, pedestrian, bicycle, and background. This plugin shows the segmentation overlay on video output.

Features

The following table summarizes the features of the plugin.

Gst-nvsegvisual plugin features

Gst Properties

The following table describes the Gst properties of the Gst-nvsegvisual plugin.

Gst-nvsegvisual plugin gst properties