Using multiple models on the same hailo 8 module

Hello,

I am working with a setup that includes a Hailo-8 AI accelerator module and I have a few questions regarding its capabilities:

Can I run two models, one object detection and one segmentation, on the same hailo-8 module?

If not, is it possible to run two detection models at the same time?

Thanks!

Yes, you can. The HailoRT scheduler allows you to run pipelines with multiple models.

Our Tappas License Plate Recognition examples demonstrates this:

GitHub - Tappas - License plate recognition

You can also try this without an application using the HailoRT CLI run2 command. It allows you to run multiple networks and measure the overall performance you can achieve. The general syntax is something like this:

hailortcli run2 set-net model_1.hef set-net model_2.hef

For each network you can add parameters for instance to set framerate and batch-size.

You can use the HailoRT CLI monitor to see the networks running.

Thank you very much, so can I run 1 yolov11n model for object detection task and 1 yolov11n-seg model for instance segmentation task, but the input must be from a camera like pi camera, I tried to run it but got error related to gstreamer.
Can you tell me the direction I should do?

Hi @DevPhamPham
You can use PySDK for these type of use cases: Running multiple models independently

Hi @DevPhamPham ,

You can check out: hailo-rpi5-examples/community_projects/detection_cropper at main · hailo-ai/hailo-rpi5-examples · GitHub

This application demonstrates a cascading TAPPAS pipeline using the Hailo cropper pipeline element. The pipeline performs the following steps:

  1. Object detection using the YOLO network.
  2. If a person is detected, the bounding box of the detected person is cropped.
  3. The cropped bounding box is sent to a depth estimation network.

For more details, please have a look here: tappas/README.rst at master · hailo-ai/tappas · GitHub

1 Like