- 8
- 47 022
Jung
เข้าร่วมเมื่อ 27 พ.ย. 2023
I like learning stuff.
I will try some random stuff and put video up if I think i could be useful for someone :)
I will try some random stuff and put video up if I think i could be useful for someone :)
Raspberry Pi 5 - Ep07 - UPS Module 3S
Trying UPS (Uninterruptible Power Supply) and bit of outdoor action (just simple video recording with RPi)
then trying Yolov8 car plate detection for a test.
Items in the video:
www.waveshare.com/wiki/UPS_Module_3S
Checking battery status: 01:53
How to configure RPi: 03:21
Outdoor action: 06:40
#raspberrypi5 #yolov8 #python #ups
then trying Yolov8 car plate detection for a test.
Items in the video:
www.waveshare.com/wiki/UPS_Module_3S
Checking battery status: 01:53
How to configure RPi: 03:21
Outdoor action: 06:40
#raspberrypi5 #yolov8 #python #ups
มุมมอง: 2 148
วีดีโอ
Raspberry Pi 5 -Ep06 - Coral TPU / hatdrive / Yolov8 / Python 3.11
มุมมอง 9K4 หลายเดือนก่อน
Running Coral TPU examples in Python 3.9(docker), Python 3.11(conda). Also experimenting what resolution of yolov8 we can run in the TPU. Items in the video: coral.ai/products/m2-accelerator-bm pineberrypi.com/products/hatdrive-bottom-2230-2242-2280-for-rpi5 Coral device setup guide: www.jeffgeerling.com/blog/2023/pcie-coral-tpu-finally-works-on-raspberry-pi-5 Really great article! Yolov8 test ...
Raspberry Pi 5 - Ep05 - Animal Surveillance Camera? - Python/OpenCV/Yolov8
มุมมอง 1.5K5 หลายเดือนก่อน
Raspberry Pi 5 experiment with yolov8 model (640x640). If a cat, dog or birds appears in the frame, video recording starts. Also you can stream the video from Raspberry Pi to your PC. Birds: 2:06 Birds2: 3:44 Cat: 4:07 Repo: github.com/JungLearnBot/RPi5_yolov8/blob/main/Readme.RPi5.cpu.picam.qt.animal_recorder.md github.com/JungLearnBot/RPi5_yolov8/blob/main/animal_recorder.py #raspberrypi5 #yo...
Raspberry Pi 5 - Ep04 - Raspberry Pi Camera Module 3 - Python/OpenCV/Qt5
มุมมอง 23K6 หลายเดือนก่อน
Running Raspberry Pi Camera Module 3 in python with OpenCV and Qt5 Items in the video: www.raspberrypi.com/products/camera-module-3/ www.raspberrypi.com/products/camera-cable/ www.raspberrypi.com/products/27w-power-supply/ Code: github.com/JungLearnBot/RPi5_yolov8/ github.com/JungLearnBot/RPi5_yolov8/blob/main/Readme.RPi5.cpu.picam.qt.md Putting things together - 0:0 Trouble shooting the pi cam...
Yolov8 Training Custom Resolution
มุมมอง 5556 หลายเดือนก่อน
This video shows how to train yolov8 in windows with Nvidia GPU. This is how I trained 320x320 model to run on Raspberry Pi 5 in th-cam.com/video/ZebczOt90mU/w-d-xo.html #yolov8 #python #computervision repo: github.com/JungLearnBot/RPi5_yolov8
Raspberry Pi 5 - Ep03 - Object Detection/Yolov8/CPU
มุมมอง 7K6 หลายเดือนก่อน
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: github.com/JungLearnBot/RPi5_yolov8/tree/main Edit: Now I made a video of how I trained a 320x320 yolov8 model: th-cam.com/video/n7_QIoCOzaE/w-d-xo.html Clone the...
Raspberry Pi 5 - Ep02 - Conda/OpenCV/WebCam
มุมมอง 2.1K6 หลายเดือนก่อน
Installing Mini Conda for future python projects. For a test, running opencv script to access a web cam. #raspberrypi5 #python #anaconda3 Commands used: // install mini conda mkdir -p ~/miniconda3 wget repo.anaconda.com/miniconda/Miniconda3-latest-Linux-aarch64.sh -O ~/miniconda3/miniconda.sh bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3 ~/miniconda3/bin/conda init bash //optional clean ...
Raspberry Pi 5 - Ep01 - Getting Started
มุมมอง 1.1K6 หลายเดือนก่อน
My First attempt to learn Raspberry Pi. Also my first attempt to do a youtube video :) Items in the video: www.raspberrypi.com/products/raspberry-pi-5 www.raspberrypi.com/products/active-cooler OS: Raspberry Pi OS (64-bit) #raspberrypi Pi #raspberrypi5 #iot
Hi Jung. Great video. By following it and the docs in your github repo I was able to get the PyCoral classification example working with the Coral USB Accelerator that I'm using with my Pi 5. I was wondering if you got the PyCoral detection example to run? When I try it I get a segmentation fault with no error message. It runs OK on the CPU if I specify a non-edgeTPU model. I don't know if it's because the model is too large (but then why do they include it) or due to changed quantization requirements for TF2. I'm yet to try Yolov8 192 - maybe I'll have more luck with that.
can you try yolov10 and inference time for each v10 model
Maybe you can use Google Coral?
⚠️ 📢 Question. Im using roboflow on my surface pro 9 and steam deck my model is getting 5fps with this be faster? I need at least 30fps
Can i have a look for the script?
Hii May I have your contact number just to ask for your help, I m currently facing an issue about the webcam object detection is low in fps, so does lower resolution provide higher fps in Raspberry Pi5 ?
Hi bro how to i connect with you i need some help in raspberry pi python camera coding 😢
Key moments 1:46 and 5:35 Good video, but could have used some editing :)
is it compatible with raspberry pi 4 ?
No the AI hat only works for the Raspberry PI 5.
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?
can i get ur number or email that i cant contact with u
why does this one "bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3" not work for me? it just say "bash: /home/hin/miniconda3/miniconda.sh: No such file or directory" i follow u everystep :(
so u dont need to use every code the github provide or how do i know when to use all or just some of it
Hi. If you see the description. It says which files (readme) are for this video
1:16 Thanks man for highlighting the fact that those camera modules don't work with the RasPi 5, at least a video on that XD but I couldn't find it before on my search XD weird, but ok I already made them ordered the cables of 22 to 15 pins... Btw and FYI: 1:16 yeah that power supply is much better, but I have no warning at all with my power supply charger of my Samsung Galaxy S9+, well it's a new charger I have because the previous one was too old and died (I think from a fall) and this is a simple power supply of 10 W 5V and 15 W 9V, I didn't have any issues at all, it just heats a bit while connected to the RasPi5. Just to let know others in case they cannot get the official power supply. And yeah that is good without the UK connector XD I also made them order (kind of bosses of the project) one and it comes the same just with rounded connectors for the outlets. :)) I'll keep taking a look at your video. Like 274! 2:04 hopefully no man, is a fact it will XD haha they are designed for it ;) 3:54 haha no worries man it happens, but yeah indeed the black thing goes from the side of the supply power, the connectors facing the Ethernet connector. 4:54 they don't blow up, maybe heating up or turning off but not blowing up XD
What do you thing of this example with Rpi5 and coral? It seems that it's enough to detect intrusions (Frigate): th-cam.com/video/so8_h0EKPWU/w-d-xo.html
Hi! Thank you for your time and effort making these videos, I have the pi5 and cam module 3, I prepared the environment according to your description on github, and when I try running "python main_picam.py --camera_test" I get the following error "File "/home/berry/Downloads/RPi5_yolov8-main/main_picam.py", line 8, in <module> from picamera2 import Picamera2 ModuleNotFoundError: No module named 'picamera2'" You mentioned about manually copying the picamera2 files to the conda environment, which I did as well. Seems like I am missing the library? I am a beginner using raspberry pi and python.
Did you pip install pycamera2? pypi.org/project/picamera2/
Sir, how can i contact you? I have some doubts regarding my raspberrypi opencv project.
Hello Mr.. great work Do you have a link to install ultralytics on RPi 4 ?
I am currently trying to do this, yet I have a Raspberry Pi 4. I am assuming you are using Bookworm OS ? In theory it should also work for the RB4 ?
Yes it would since it is just python code. Thou you probably have to change code a bit if you use different camera.
@@JungLearnBot I'm using a PiCamera V2 which someone in the comments already confirmed it work. Will work on this later today so I'll post if it indeed works. Ty for taking the time to doing this, kind of sad that this is so buggy considering there is a slot on the boards for these specific cameras. I'll also try USB to see the difference on the approaches.
Good job. Make a video to run llm into raspberry pi 5 with coral edge tpu
I wouldn't have figured out how to connect the camera if it wasn't for this video. I thought you had to use some other kind of other connector or electrical tape to get the original cable to work with the adapter 😅
Glad it helped :)
hello, i don't understand about how to use my custom model here
If it is yolo v8 model, you just need to specify the model name as argument
hello awesome tutorial but I am having a problem so I want to ask you if you help me user@raspberrypi:~ $ bash ~/miniconda3/miniconda.sh -b -u -p ~/miniconda3 PREFIX=/home/user/miniconda3 Unpacking payload ... /home/user/miniconda3/miniconda.sh: line 1258: /home/user/miniconda3/_conda: cannot execute: required file not found user@raspberrypi:~ $ I dont know what to do with it
Excellent!
Many thanks!
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?
Hmm. Have you checked what fps you get for onnx format? Also i would double check what is fps of the input camera first without any other operation
hi jung , I got more than 100 fps for input camera and for onnx I got less than 2 fps. @@JungLearnBot
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
As it is just python code, most of the code should be okay for raspberries pi 4. Only bits that could be different is the code related to camera and it would depend on what camera module you have
Very helpful video thank you!
Glad it was helpful!
I tried running it, but I got always blank output. Do you have something to help? Btw: GREAT!
Hi @Willie, what error msg do you get? do you see coral tpu: /dev/apex_0 When you do run: ls /dev/apex_0 ? Or maybe there is problem with camera? Let me know what errors you get plz
I opened issue on GitHub, so let's continue conversation there
I replied your comment in github :)
I'm experiencing the same problem with the cable. Will the camera cable meant for the Raspberry Pi 0 function properly with the Raspberry Pi 5? I've come across a cable identical to the one you're using, but it's labeled as 160FPC instead of 150. What does this number even mean? And more importantly, will it be compatible?
I am using raspberry pi zero cable and it seems to work without too much issue. I heard they are pretty much the same but not sure about the number, sorry
just bought two of them,. The number represents the length of the cable.@@JungLearnBot
Thanks for sharing this video. Curious why you didn't convert your model to TFLite and quantize it, in order to get faster performance. Please share your thoughts!
Thank you for your comment, @oliverexcellent. I tried quantised model as well but I found its accuracy was bit lower than the full version (which was expected as it is quantized ver). I would have used quantized version if I could run it in realtime at 640x480. I didn't want to use lower the resolution as I was trying to detect animals appear relatively small due to camera setup (it would been slightly different story if it was about detecting animal in very close range). One of idea for this setup was to call detection less frequently with relatively high resolution (640x480) and best detection quality. so yes, it was kinda on purpose. but using quantised model would work as well. hope this gave you some context
Wow. I'm surprised at your results as well. I had high hopes for its output. I just purchased the double TPU and just found out that the hat can only handle one TPU. I hope I find out differently here
@mikew2883 Sorry to hear dual TPU didnt work for you. I was also thinking about the dual TPU at some point as well but yes, it seems you need a different hat and it seems like you have to segment a model to utilize multiple coral TPU. coral.ai/docs/edgetpu/pipeline/#overview
thank you this was extremely helpful, tested on a pi cam v2
Great to hear!
Were you using a Raspberry Pi 5 with the pi camera module 2?
@@gregorymoore7617yup!
@@gregorymoore7617yup!
@@gregorymoore7617yup!
Hi … can i use this code also on a pi zero 2w … sorry but i’m new to this raspberry
I think it should if you are using camera module 3 and bookworm os but not 100% sure as i havnt tested in that configuration.
Does camera module rev1.3 work in rpi 5?
Sorry. I haven’t tried rev1.3 so I’m not sure
@@JungLearnBot sorry its working ,as its in the video wrongly connected camera
Nice!
can you please install mediapipe on rpi 5.
@jaysoni9568 have you tried pip install? I just tried following and mediapipe seem to install without issue conda create -n mediapipe python=3.11 conda activate mediapipe pip install mediapipe
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 what is the exact command did you run? and what was the output? saved_model folder gets generated automatically in the process of export. maybe some dependencies are not installed correctly?
@@JungLearnBot i use the export program exactly as you shown. I will look more info dependencies.
Can raspberry pi 4 cameras be used with raspberry pi5? Or is there some specific cameras for Raspberry pi 5?
Pi cameras modules can be used for both RPi4 and RPi5. (At least pi camera module 3). Thou you need slightly different cable for RPi5. www.raspberrypi.com/products/camera-cable/ If you use RPi4, you can use the cable which comes with the camera.
Did you used Raspberry pi 5 4g or 8gb version?
Im using 8gb ver atm
Jung, it is really a great job. I think the most important thing is whether the performance at the 192 resolution can detect things correctly. It detected person correctly at 30 fps at least. I still see its potential.
Hi Jung, I use the export_models.py to convert my custom weights to tflite. But for some reason, it keeps showing "person", which is the default yolo weights. Can please advise? My OD is to detect holes
something has changed, I can't install picam2 into the conda environment. it stops at "Preparing metadata (setup.py) ... error " and "You need to install libcap development headers to build this module". any idea?
@robertkloetzer1655, sorry. I haven't met that problem yet. but according to: github.com/raspberrypi/picamera2/issues/294 sudo apt install -y python3-libcamera python3-kms++ libcap-dev can solve the problem. hopefully this fixes your issue :)
Yes, it works. thank you very much, great work @@JungLearnBot
Man, is there is any problem if I am using raspberry pi 4 power supply of 15 W
@jaiswal_stock, I think so, as the official power supply is 15w. www.raspberrypi.com/products/type-c-power-supply/ Only Raspberry pi 5 needs special one: www.raspberrypi.com/products/27w-power-supply/
Dear Jung, I decided to buy pi 5 and camera module v3. Will i have any problem, if i use python v 3.12 ?
I can only make camera module 3 on python 3.11 and it is causing me lots of pain. I havent tested on 3.12. I tried to make it work on 3.10 and 3.9 but failed.
Wow...thanks for tutorial
Welcome 😊
I had the same problem 3 weeks ago with the cable
Ya. This is annoying. I wish there was some description about this in pi camera module. I guess RP5 came out later than the camera and they did not bother updating the description much.
man, u saved our project, thank u
Glad I could help :D
I am trying to connect a Waveshare Model 3S (UPS) to my Rasberry Pi 4B. I am using 3 18650 batteries to power the UPS. I have checked these batteries and they are 100% good (output 4V per each). No power output when the battery is connected, nothing happens when I connect it to the Raspberry Pi 4. Can someone suggest what I should do to identify the problem? p.s thank you for the video!
Hi. @mishasavchuk4096, sorry to hear the UPS doesn't work for you at the moment. For me, I didn't need to do anything special but put batteries in and the module worked. As I am just a software engineer, I don't know much about electronics so only advice I can give you is to check if batteries are in the correct way and metal bits are correctly touching, which I am sure you checked. Apparently if you put batteries in the wrong way you will see red led but I haven't tested it. As it is supposed to work if the batteries are inserted correctly, maybe it could be a faulty unit? Maybe you need to talk to the seller.
no lo voltage warning? i already tried a UPS hat for it and it worked but kept getting the low voltage error. Thank you for the video, i will try this setup!
Thank you for the comment, @javicar31896. I do get low voltage warnings time to time but not always. The documentation says it supports 5V/5A so I thought it was due to the 18650 batteries I used or maybe my one was bit funny. Now I know others get this too! Thank you for sharing your comment :)
I've been following along, but where did coral_test/ come from? I don't have anything in my projects folder.
Hi, @bernardkim6669. Sorry for the confusion. 'coral_test' folder is not part of the git repo. It is just a folder in my local directory where I checked out "github.com/google-coral/pycoral/tree/master/examples" to verify if my local setup was correctly working. (but keep in mind that repo is very big so its up to you check out. I only checked out because I tried to build pycoral myself at the beginning.) If you got the docker version working, maybe just try to run (or use part of the code) "github.com/JungLearnBot/RPi5_yolov8/blob/main/main_picam_coral_tpu.py" This contains code of tpu + picam and it easy pretty easy to change to use opencv camera. My other examples (e.g main.py) contains code to load images from opencv(webcam). Hope this helps and if you hit any other issues, feel free to ask anytime please :)
@@JungLearnBot
@@JungLearnBot If so, does that mean I don't need to use a docker image? Can I just jump straight to installing pycoral?
@bernardkim6669 It depends on what you want. If you just want to confirm your device is working, docker image is bit easier since it requires less steps to install (install docker/build image/run image). But the problem of this is that now you have to use python 3.9 This is fine if you just use webcam. you may just run everything with docker. But in my case, I wanted to use pi camera module 3 which I must use python 3.11. So I went thru all the steps to create a conda environment and install specific version of libraries.