Tutorial of AI Kit with Raspberry Pi 5 about YOLOv8n object detection | Seeed Studio Wiki

[giscus[bot]]

Tutorial of AI Kit with Raspberry Pi 5 about YOLOv8n object detection | Seeed Studio Wiki

This wiki demonstrates how to use yolov8n to do object detection with ai kit on raspberry pi5 from traing to deploying

https://wiki.seeedstudio.com/tutorial_of_ai_kit_with_raspberrypi5_about_yolov8n_object_detection/

[Atzingen]

Not clear on how to do the "hailomz parse ..." part with a custon yolo8 dataset. Does the dataset realy have to be converted to TFRecords and if so, how to use create_coco_tfrecord.py for this ?

[LJ-Hao]

You dont have to conver dataset to TFRecords, but TFRecord will accelate training. You can check this link:Car Detection and Tracking System with RPi Hailo Al Kit for using npy dataset.

[wb666greene]

Any reason you can't just use the Ultralytics pre-trained COCO coefficients instead of redoing it? Also the "nano" model is fast but it doesn't give very good results. The "small" is accurate enough to be useful, but the "medium" is the best trade-off for speed-accuracy in my experience. I started back in the days of Ultralytics YOLOv5.

[LJ-Hao]

You are right, however, if you want to apply the model to a specific scenario, retraining the pre-trained model is necessary. I would like to demonstrate this process.

[KunalNN]

We are having issues doing it with YOLO-v8-pose. any guidience!

[dje666]

Are there any missing steps in this tutorial?

$ python3 -m pip install -v -e .

This fails with error: externally-managed-environment......

I create and activate a new virtual environment, but the get warnings about not being in the Hailo virtual environment....?

A few more instructions would be appreciated.

TTFN,

DJE666