Rpi5 picamera2 detection + tracking

General
1

Hey, it’s cool to see PiCamera2 support for Hailo with examples!

Is there an option to use tracking detection? Do you have any examples for that, or do we need to implement our own SORT algorithm? Or is it built-in, or planned for future releases? If so, how long might we need to wait?

Thanks

2

Hey @DimaR

Yes, there is an example of how we implement tracking detection. You can find it in the following code repository:

Regards

3

Hello, I have rpi5+Hailo8, I check the link and I face issue with run the detection and tracking python code because it depends on hailo_platform which isn’t supported by rpi as I understand.

can you help me please

4

Hey @aameralshamy3,

Welcome to the Hailo Community!

Hailo is supported on Raspberry Pi. To install it, simply run the following command:

sudo apt install hailo-all

For more information, please refer to the installation guide: https://github.com/Hailo-ai/hailo-rpi5-examples/blob/main/doc/install-raspberry-pi5.md

To verify the installation, open a Python terminal and try importing the hailo or hailo_platform module:

python
import hailo

or

python
import hailo_platform

If the import succeeds without any errors, Hailo is properly installed on your Raspberry Pi.

Best Regards,
Omria

1 Like
6

Hi, thank you for your help, I have very good results.
May I ask you how to set the batch size? because I need faster inference, I try modefy the batch_size: int = 1 to batch_size: int = 8 lines in utils.py file:

class HailoAsyncInference:
    def __init__(
        self, hef_path: str, input_queue: queue.Queue,
        output_queue: queue.Queue, batch_size: int = 8,
        input_type: Optional[str] = None, output_type: Optional[Dict[str, str]] = None,
        send_original_frame: bool = False) -> None:

but nothing changes.
Note:
when use: detection_pipeline → GStreamerDetectionApp, I can easily set the batch size and increase the FPS and inference speed.

Thanks

7

Hello,
can anyone help me please.