Welcome to Hailo-8 AI HAT + repository

6.Segmented application

The chunking process demonstrates how each frame can be divided into multiple slices, each processed independently by a HailoNet component. This method is particularly effective for detecting small objects in high-resolution frames.

1. Detecting Objects in Video

Default Model: hailo_yolov8n_4_classes_vga - Optimized for aerial object detection

Default Video: tiling_visdrone_720p.mp4 - Aerial video containing small objects

Application Scenarios: Ideal for demonstrating small object detection in high-resolution frames

For scenes with objects of varying sizes, you can enable multi-scale chunking using the --multi-scale flag.

Terminal Input,

cd hailo-apps

Then enter the command to enter the virtual environment.

xxxxxxxxxx
source setup_env.sh  

Enter the command. Run the video inspection.

xxxxxxxxxx
hailo-tiling

To close the window, press Ctrl+C.

2. Real-time Camera Detection

In the virtual terminal environment, enter the following command. By default, it will automatically detect available USB cameras (if multiple cameras are connected, it will use the first one detected). You can change the input source using the --input flag, for example, videox, then the input source would be:

--input /dev/videox (replace x with the actual device number).

xxxxxxxxxx
hailo-tiling --input usb --multi-scale        #USB camera
hailo-tiling --input rpi --multi-scale          #CSI camera

If the camera is not displaying correctly, you can check which video devices are available by running the following command:

xxxxxxxxxx
ls /dev/video*

You can test if the camera is working properly by running the following command:

xxxxxxxxxx
ffplay -f v4l2 /dev/video0

If you encounter an error, try a different device, such as /dev/video2.

For more information, please refer to ：hailo-apps/hailo_apps/python/pipeline_apps/tiling/README.md at main · hailo-ai/hailo-apps · GitHub