1. Pandora
2. Pandora Tutorials
3. 7. Vision Advanced
4. 7-4 Four types of Detection (Traffic)

7-4 Four types of Detection (Traffic)

Learning Objectives

We run DeepStream-app (the reference application) using a single 1080p H.264 video file as input. Through two MultiURI sources, we replicate 30 instances of the decode + inference pipeline and display the results in a 6×5 grid layout. The model detects four object classes appearing in the video—car, bicycle, person, and road_sign—and overlays bounding boxes on them for display.

Run the example

The first run requires TensorRT optimization, which takes about 10 minutes. If the container is deleted and a new one is created, TensorRT optimization will need to be performed again.

During the process, the following WARNING messages (as shown in the image) may appear. These indicate that the model is being optimized, so please do not interrupt.

Results:

Reference:

Quickstart Guide — DeepStream documentation