gstreamer detection on pi5 from tappas

General
1

hello , trying to run tappas/apps/h8/gstreamer/raspberrypi/detection on pi5 by the next approach and have some questions:

(1) that trigger the resource download and building some staff but the app inside raspberry folder (./detection.sh) calling tappas/apps/h8/gstreamer/libs/post_processes/libyolo_hailortpp_post.so which is a x86 elf not arm. shouldn’t “target-platform rpi” build that for the right arch?

bypassed (1) by override that so with hailo-rpi5-examples/resources/libyolo_hailortpp_postprocess.so but :

(2) running the ./detection.sh from the raspberrypi tappas apps complain about “no element hailofilter” while gst-inspect-1.0 was installed as instructed on tappas/docs/installation/manual-install.rst at master · hailo-ai/tappas · GitHub ;

(3) there is any better way to use tappas apps (not just the core) on the pi5?

Thanks a lot for a great device;

2

Hello @nanodev,

Welcome to the Hailo Community!

We currently do not recommend using TAPPAS on the Raspberry Pi, as we are in the process of transitioning all TAPPAS pipelines to the Hailo RPi examples repository for easier use and better compatibility.

Here’s what we recommend:

  1. Install the hailo-all package:
    This package includes the Hailo driver, HailoRT, TAPPAS Core, and pyHailoRT, providing all the necessary components for running pipelines on the Pi.

  2. Use the Detection Pipeline from the Hailo RPi Examples:
    The detection pipeline is already set up in the Hailo RPi examples repository. After installing the hailo-all package, follow the steps provided in this repository to run the detection pipeline:
    Hailo RPi5 Examples

This setup will ensure a smoother experience and compatibility on the Raspberry Pi. Let us know if you encounter any issues!

Best Regards,
Omria

3

Thank you for the response!

I’ve explored the Hailo RPi5 Examples repository, particularly the basic_pipelines, and as the name suggests, these examples are quite basic. In contrast, the TAPAS repository offers a significantly richer starting point. Its focus on:

Accelerating time-to-market by reducing development and deployment efforts.
Providing a foundation for customers to fine-tune their applications.

I noticed some advanced use cases in the TAPAS repository, such as the multi-model pipeline with tracking in tappas/apps/h8/gstreamer/general/license_plate_recognition. This is an impressive example, and it would be amazing to have something equally sophisticated running on the RPi5.

Until the porting effort you mentioned is completed for the Hailo RPi Examples repository, are there any options or workarounds to build and run the TAPAS repository on the RPi5 in the meantime?

Thanks again for providing such a fantastic product!

4

… will also be glad to know if there is an option to add or install gstreamer pluigns for the pi5 like the hailotools for example that is seen under gst-inspect-1.0 of the suite…

Thanks

5

Hi @nanodev
We developed a python SDK to make it easier for users to develop applications using Hailo+RPi. Please see Simplifying Edge AI Development with DeGirum PySDK and Hailo and let us know if we can help you with anything specific.

6

… found a solution using the post form Face Detection on RPI5 using Hailo8L | by Sanjoy | Medium that used the tappas_v3.29.1 from the dw-center (dont sure why there is no tappas-arm64 for latest version);