hailo-apps Issue with using hailo-seg with a different model than the default

Hello,

I have been trying to get started with the Hailo8 on an RPi5, but I am running into problems with using the instance segmentation example “hailo-seg”. It works fine with the default settings (ex: hailo-seg -i rpi), but if I try to change the HEF model from the default (ex: hailo-seg -i rpi -n yolov5l_seg), I get an error message that the HEF version is not supported:

(venv_hailo_apps) mark@rpi5-hailo:~/hailo-apps $ hailo-seg --list-models

============================================================
Available models for: instance_segmentation (hailo8)
============================================================

📦 Default Models:
   • yolov5m_seg

📚 Extra Models:
   • yolov5l_seg
   • yolov5n_seg
   • yolov5s_seg
   • yolov8m_seg
   • yolov8n_seg
   • yolov8s_seg
   • yolov5m_seg_with_nms
   • fast_sam_s

============================================================
Total: 1 default, 8 extra

Usage: --hef-path <model_name>
       Model will be auto-downloaded if not found locally.

(venv_hailo_apps) mark@rpi5-hailo:~/hailo-apps $ hailo-seg -i rpi -n yolov5l_seg
INFO | common.core | All required environment variables loaded successfully.

⚠️  WARNING: Model 'yolov5l_seg' is not downloaded.
   Downloading model for instance_segmentation/hailo8...
   This may take a while depending on your internet connection.

Downloading resource: yolov5l_seg (type: model)
✓ URL valid: https://hailo-model-zoo.s3.eu-west-2.amazonaws.com/ModelZoo/Compiled/v2.17.0/hailo8/yolov5l_seg.hef
  Status: 200
  Size: 47980499 bytes
[=======================================-] 99% (45.30/45.76 MB)
INFO | common.core | Model downloaded successfully: /usr/local/hailo/resources/models/hailo8/yolov5l_seg.hef
INFO | instance_segmentation.instance_segmentation_pipeline | HEF path: /usr/local/hailo/resources/models/hailo8/yolov5l_seg.hef
ERROR | instance_segmentation.instance_segmentation_pipeline | Unsupported HEF version: yolov5l_seg.hef
Traceback (most recent call last):
  File "/home/mark/hailo-apps/venv_hailo_apps/bin/hailo-seg", line 7, in <module>
    sys.exit(main())
             ~~~~^^
  File "/home/mark/hailo-apps/hailo_apps/python/pipeline_apps/instance_segmentation/instance_segmentation.py", line 149, in main
    app = GStreamerInstanceSegmentationApp(app_callback, user_data)
  File "/home/mark/hailo-apps/hailo_apps/python/pipeline_apps/instance_segmentation/instance_segmentation_pipeline.py", line 115, in __init__
    raise ValueError(
        "HEF version not supported; please provide a compatible segmentation HEF or config file."
    )
ValueError: HEF version not supported; please provide a compatible segmentation HEF or config file.
(venv_hailo_apps) mark@rpi5-hailo:~/hailo-apps $ 

I saw today’s post that hailo-apps was updated, so I did the following to update:

$ deactivate
$ sudo apt update
$ sudo apt full-upgrade
$ cd ~/hailo-apps
$ git pull origin main
$ sudo ./install.sh
$ source setup_env.sh

That seemed to resolve a different issue I was having with the python/standalone_apps/instance_segmentation not working with “-i rpi” or “-i camera”, but did not resolve the above.

I tried the “hailo-pose” example and it works if I select a different model (hailo-pose -i rpi --show-fps -n yolov8s_pose).

My hardware:

RPi5 (16GB w/ WIFI), with AI HAT+ Hailo8, from CANAKIT (with 128GB microSD)
Canakit USB supply (5V @ 5.1A capable)
Raspberri Pi Camera module 3 (Wide angle version)
USB Keyboard and mouse

I am using a fresh image of Raspberry PI OS (64-bit) (Debian Trixie) that I made yesterday (Jan 27, 2026). The only items I have installed are the hailo apps, and VSCode. I installed VSCode after the above issues to have a better interface for trying to debug the issue.

Thanks @Mark_R We are looking into this.