Hailo TAPPAS - Optimized Execution of Video-Processing Pipelines

.. |gstreamer| image:: https://img.shields.io/badge/gstreamer-1.16%20%7C%201.18%20%7C%201.20-blue :target: https://gstreamer.freedesktop.org/ :alt: Gstreamer 1.16 | 1.18 | 1.20 :width: 150 :height: 20

.. |hailort| image:: https://img.shields.io/badge/HailoRT-4.24.0%20%7C%205.5.0-green :target: https://github.com/hailo-ai/hailort :alt: HailoRT 4.24.0 | 5.5.0 :height: 20

.. |license| image:: https://img.shields.io/badge/License-LGPL--2.1--or--later-green :target: https://github.com/hailo-ai/tappas/blob/master/LICENSE :alt: License: LGPL-2.1-or-later :height: 20

.. |check_mark| image:: ./resources/check_mark.png :width: 20 :align: middle

.. image:: ./resources/github_Tappas_Mar24.jpg :height: 300 :width: 600 :align: center

|gstreamer| |hailort| |license|

---

Overview

TAPPAS is Hailo's infrastructure for building applications, implementing pipeline elements and pre-trained AI tasks.

Hailo applications are now maintained at this repository <https://github.com/hailo-ai/hailo-apps-infra>_.

Demonstrating Hailo's system integration scenario of specific use cases on predefined systems (software and Hardware platforms). It can be used for evaluations, reference code and demos:

• Accelerating time to market by reducing development time and deployment effort
• Simplifying integration with Hailo’s runtime SW stack
• Providing a starting point for customers to fine-tune their applications

---

Getting Started with Hailo-8 And Hailo-10H

Prerequisites ^^^^^^^^^^^^^

• Hailo-8 or Hailo-10H device
• HailoRT PCIe driver installed
• At least 6GB's of free disk space

.. note:: This version is compatible with HailoRT v4.24.0 for Hailo-8 devices, and with HailoRT v5.5.0 for Hailo-10H devices.

Installation ^^^^^^^^^^^^

.. list-table:: :header-rows: 1

• • Option
  • Instructions
  • Supported OS
• • Hailo SW Suite*
  • SW Suite Install guide <docs/installation/sw-suite-install.rst>_
  • Ubuntu x86 24.04, Ubuntu x86 22.04
• • Manual install
  • Manual install guide <docs/installation/manual-install.rst>_
  • Ubuntu x86 24.04, Ubuntu x86 22.04, Ubuntu aarch64 20.04
• • Yocto installation
  • Read more about Yocto installation <docs/installation/yocto.rst>_
  • Yocto supported BSP's
• • Raspberry Pi 5 installation
  • Read more about Raspberry Pi 5 installation <https://github.com/hailo-ai/hailo-rpi5-examples/blob/main/doc/install-raspberry-pi5.md>_
  • Raspberry Pi OS

* It is recommended to start your development journey by first installing the Hailo SW Suite

Documentation ^^^^^^^^^^^^^

• Framework architecture and elements documentation <docs/TAPPAS_architecture.rst>_
• Guide to writing your own C++ postprocess element <docs/write_your_own_application/write-your-own-postprocess.rst>_
• Debugging and profiling performance <docs/write_your_own_application/debugging.rst>_
• Cross compile <tools/cross_compiler/README.rst>_ - A guide for cross-compiling

---

Getting Started with Hailo-15

TAPPAS is now released separately for Hailo-8 and Hailo-10H, for Hailo-15 please refer to https://github.com/hailo-ai/hailo-camera-apps.

For a quick start with Hailo-15, please refer to the Vision Processor Software Package documentation section in Hailo's Developer Zone <https://hailo.ai/developer-zone/documentation/>_.

---

Example Applications Built with TAPPAS

TAPPAS includes a single-stream object detection pipeline <apps/detection/README.rst>_ built on top of GStreamer. These example application is part of the Hailo AI Software Suite.

Hailo offers an additional set of Application Code Examples <https://github.com/hailo-ai/Hailo-Application-Code-Examples>. For the Raspberry Pi 5 applications, go to Hailo Raspberry Pi 5 Examples <https://github.com/hailo-ai/hailo-rpi5-examples>.

.. important:: * Example application utilize both the host (for non-neural tasks) and the Neural-Network Core (for neural-networks inference), therefore performance results are affected by the host. * This application example does not include any architecture-specific accelerator usage, and therefore will provide the easiest way to run an application, but with sub-optimal performance.

.. note:: Running application examples requires a direct connection to a monitor.

---

Support

If you need support, please post your question on our Hailo community Forum <https://community.hailo.ai/>_ for assistance.

Contact information is available at hailo.ai <https://hailo.ai/contact-us/>_.

---

Changelog

v5.3.1 (June 2026)

• Deprecated the hailopython GStreamer element; it will be removed in a future release
• Postprocess: deduplicated common helpers (sigmoid, dequantize, NMS); unified yolo_hailortpp / mobilenet_ssd / yolo_postprocess via templates
• Overlay: added YUY2 color format support (draw_text, draw_line, draw_ellipse, blur)
• Fixed hailoaggregator / hailotileaggregator segfault on caps queries when wired directly to autovideosink without hailooverlay
• Fixed GstHailoBaseCropper caps-query abort with fakesink or multi-format peers
• Detection example: auto-detect Debian multiarch vs flat /usr/lib layout for Yocto images; added --no-display for headless runs; switched display sink to autovideosink for cross-platform display support
• This release supports both HailoRT v4.24.0 (Hailo-8/8L) and HailoRT v5.3.0 (Hailo-10H)

v5.3.0 (April 2026)

• Documentation: refactored Yocto section <docs/installation/yocto.rst>_ — removed Dunfell references, updated to Kirkstone syntax, removed deprecated recipes and i.MX warning
• Documentation: fixed outdated directory structure and paths in Writing Your Own Postprocess <docs/write_your_own_application/write-your-own-postprocess.rst>_ guide
• Documentation: fixed Hailo Objects API <docs/write_your_own_application/hailo-objects-api.rst>_ inconsistencies with source code, added HailoMask hierarchy and HailoUserMeta documentation
• Fixed get_full_percision typo — renamed to get_full_precision across source code, Python bindings, and documentation
• This release supports both HailoRT v4.23.0 (Hailo-8/8L) and HailoRT v5.3.0 (Hailo-10H)

v5.2.0 (January 2026)

• Added --static-opencv option to install.sh for static OpenCV linking.
• Added Python wheel builder for TAPPAS Python binding package
• Updated pybind11 to support NumPy 2.x
• Removed Hailo-15 support and related code
• This release supports both HailoRT v4.23.0 (Hailo-8) and HailoRT v5.2.0 (Hailo-10H)

v5.1.0 (October 2025)

• Downloader: removed redundant CLI arguments (--platform, --app-list);
• Downloader: HEF files now downloaded from model_zoo and media files from the TAPPAS bucket; removed the uploader;
• Detection app: detection.sh now supports --arch (Hailo-8/Hailo-10H);
• Models and resources: migrated model files to model_zoo; TAPPAS bucket is now used only for general MP4 files; updated resources directory structure; changed yolov5m_wo_spp_60p.hef to yolov5m_wo_spp.hef.
• Hailo-10H support: added Hailo-10H HEF downloads.
• Build and packaging: separated GCC apt installation and removed fixed GCC version; updated related documentation.
• Dependencies: updated package versions for Python 3.13 compatibility; migrated pandas to support the newer environment.
• Cleanup: removed Hailo-8 references where appropriate; removed nested directories under apps; various comment updates.
• This release supports both HailoRT v4.23.0 (Hailo-8) and HailoRT v5.1.0 (Hailo-10H)

v5.0.0 (July 2025)

• All example applications, except the object detection application, are now maintained at Hailo Applications <https://github.com/hailo-ai/hailo-apps-infra>_.
• Updated manual installation process
• Added support for Ubuntu 24.04
• Added support for Python 3.12
• This release supports both HailoRT v4.22.0 (Hailo-8) and HailoRT v5.0.0 (Hailo-10H)
• Known issue: When installing via GitHub, only Hailo-8 models are downloaded.