IMX8 Yocto - Yolov5

General
1

Hello,

I followed a procedure to use a Hailo 8 and Yolov5 on Imx8 Yocto just to test the basic version that comes with the Hailo.

My input is a webcam and my output is a screen.

The screen displays what it should, and I have frames around the objects. The problem is that there are no labels for the frames, labels such as “person”, “dog” etc …

Can you help me solve this problem please?

Here is what i modified in the basic detection.sh by Hailo :

PIPELINE=“gst-launch-1.0
v4l2src device=$input_source ! image/jpeg, format=yuv2, width=1280, height=720, framerate=30/1 !
decodebin ! videoconvert ! queue leaky=downstream max-size-buffers=30 max-size-bytes=0 max-size-time=0 !
hailonet model-path=$model_path ! queue leaky=downstream max-size-buffers=30 max-size-bytes=0 max-size-time=0 !
hailooverlay ! queue leaky=downstream max-size-buffers=30 max-size-bytes=0 max-size-time=0 !
videoconvert ! autovideosink name=halo_display sync=false text-overlay=true ${additional_parameters}”

And my json file :
{
“iou threshold”: 0.45,
“detection threshold”: 0.3,
“output activation”: “none”,
“label offset”: 1,
“max boxes”: 200,
“anchors”: [
116, 90, 156, 198, 373, 326,
30, 61, 62, 45, 59, 119,
10, 13, 16, 30, 33, 23
],
“labels”: [
“unlabeled”,
“person”,
“bicycle”,
“car”,
“motorcycle”,
“airplane”,
“bus”,
“train”,
“truck”,
“boat”,
“traffic light”,
“fire hydrant”,
“stop sign”,
“parking meter”
(and the rest of the labels)

Thank you very much.
Paramdeep

2

Hi @Paramdeep.Hardeep,

Welcome to the Hailo Community!

Let’s debug this step by step:

  1. First, let’s verify your model’s behavior on Ubuntu (assuming it’s from our Model Zoo):
  • Run a test inference on Ubuntu
  • Check if the labels match those from the Model Zoo
  • Verify the outputs are correct in a standard test run

This helps us confirm the model is working correctly outside the Yocto environment.

  1. Then, update your pipeline by adding the text overlay parameter:
hailooverlay text-overlay=true !

Let us know how these steps work for you, and we’ll help you troubleshoot further if needed!

Best regards,
Omria

3

Hello Omria, thanks for your answer. Unfortunately, text-overlay doesn’t work. And I’m using Hailo Tappas and it comes with a detection application, i don’t have a trained Model Zoo.

It detects the objects and persons but It doesn’t show the label of what the object is like “keyboard”, “mouse” etc …

Best regards,
Paramdeep

4

Hi @Paramdeep.Hardeep
May I suggest you to look at our PySDK, a python package we developed to make development easy: Simplifying Edge AI Development with DeGirum PySDK and Hailo