Custom Object Detection on Jetson Nano for 4 RTSP

we want to deetct objects ( vehicles masks and pedestrians) from 4 live CCTV feed. And document the Captured objects.

  1. Is there a pipeline which is ready to modify and use?
  2. Can Deepstream directly produce an output in csv/txt etc
  3. Is Kafka needed?

We have DeepStream SDK for this usecase. Please check

After the installation, you should see the samples in


The supported source types are listed in
DeepStream Reference Application - deepstream-app — DeepStream DeepStream Version: 5.0 documentation (

But How to i integrate a custom YOLOV5 with deep stream. What should be the changes in config file

We have demonstration for Yolo models:


You may refer to the sample and apply your model.