Raspberry Pi 5 - Ep03 - Object Detection/Yolov8/CPU

preview_player
Показать описание
Trying Yolov8(object detection) on Raspberry Pi 5.
Please note this is running without 5V/5A so the performance of the Pi is immitted. I'll test once the power supply arrives. (Currenlty just using IPad charger for the power).

Github:

Edit:
Now I made a video of how I trained a 320x320 yolov8 model:

===============================================
Clone the example repo:

Setup Conda environment:

Export yolov8n to tflite and onnx:

Demo 640x640:

Demo 320x320:

#raspberrypi5 #yolov8 #python

For more info about Yolov8:
Рекомендации по теме
Комментарии
Автор

thank you so much for the video Jung, exactly what I am looking for!

drelvenkee
Автор

Jung, nice work and a lot of info! Are you going to try to use Google Coral device like the USB Accelerator with your pi5?

clranalytics
Автор

Hello, is there a way to print out the coordinates of the bounding boxes. I'm trying to make a project using this video by trying to create my own model with the same structure as the one you used in the tutorial, but i also want to know the coordinates of the bounding boxes during the live camera. Is there a way to do that?

mantoe
Автор

I managed to have 14 to 17 fps using a tensorflow lite ssd mobilenet model, it's very good

pocketgamer
Автор

Hello Mr..
great work
Do you have a link to install ultralytics on RPi 4 ?

ghadahh
Автор

can you try yolov10 and inference time for each v10 model

rsemihkoca
Автор

Do I buy 4 gb or 8 gb for object detection only?

rahulmahatha
Автор

Did you used Raspberry pi 5 4g or 8gb version?

ranlevi
Автор

hi jung,
I got only 2 fps while using a quantized int8 tflite model with 640x640 in raspberry pi 5. I just follow tflite documentation of TensorFlow. what should i need to do additionally to increase speed?

shubhammakwana
Автор

How much FPS did you accomplish with Raspberry Pi 5? What is average inference time in ms?

mindaugasjuska
Автор

Greate video... but how can i do the say step with Raspberry Pi 4? I mean the files, i have my wont pretrain model yolov8m, and i want to do object detection using Raspberry pi4

STab-ob
Автор

Could you make a video of how to get more fps on the pi5? I've run the --debug command and only get around 7 fps, but my cpu is only at roughly 50%
Is there a way to use more cpu to get more fps?

Greenleader-bddm
Автор

Given all the news about bitcoin recently due to its price escalation, do you have any interest in creating some videos around the subject of setting up a coin nodes on a raspberry pi five? The new top and bottom hats that can hold NVME drive connected through PCIE would be a very interesting subject. It would be fun to see if you different variations of the bitcoin nodes created and demonstrated in your video. I would also think this would be the best way to drive your subscriber count higher.

ksoonsoon
Автор

Hi! @Jung I have got a lot of help from your video. But I am getting Low FPS on my yolo model running on raspberry pi 5. Can you Please upload your custom 320 tflite model so that we can get better FPS. We are doing our project and we require faster model. If you can upload it It will be very helpful.
Thanks

M.Salmanaziz
Автор

im getting this error while trying to download tensorflow

note: This error originates from a subprocess, and is likely not a problem with pip.
ERROR: Failed building wheel for h5py
Failed to build h5py
ERROR: ERROR: Failed to build installable wheels for some pyproject.toml based projects (h5py)

eliemutia
Автор

hello, i don't understand about how to use my custom model here

SwethaAnbalagan-sy
Автор

Hey, thank you so much for your video, i cant export to tflite due to saved_model non existent, do you have a solutiion for this?

reeeeoo
Автор

Raspberri pi 5 have a pcie port for ai accelerator card

maxmed
Автор

Why can't you just import PiCamera?

csr-vhzv