Multi-stream Object Detection with Hailo8 (OpenCV, no GStreamer)

General
,
1

Hi,
I’m working with Hailo8 and currently testing object detection using the example code here:
Hailo-Application-Code-Examples – object_detection (C++)
My requirement is to handle multiple input streams (e.g., several video feeds) and run inference on all of them. I would like to manage the streams in code and pass them to Hailo for inference.

Is there any example code available for handling multiple video streams with OpenCV (without using a GStreamer pipeline)?

Thanks!

2

Hi we do not have a multistream example without gstreamer.

We will release soon a new python example for multistream but it will be also use gstreamer.

Gstreamer allows us to utilize zero copy, multithreading, video operations and additional tricks to reach high performance. You can modify the code to use multiple stream and send each one to inference in a round robin fashion if it is good enough for you.

If you want multiple streams in real time I suggest using our gstreamer based example.

3

Hi @giladn
Is there a sample C/C++ application that uses a single .hef file while processing multiple input streams (with or without GStreamer)?
We’re trying to adapt the current example to accept multiple video sources, so a reference showing how to handle and pass frames from different streams through Hailo would be very helpful.

4

Hi, We do not have a C/CPP example for multisource. We will release soon a python + Gstreamer app. It can be modified easily to use CPP on top of Gstreamer instead of python.