Raspberry Pi Object Detection Tutorial
HTML-код
- Опубликовано: 3 авг 2024
- Here's how you can make your Raspberry Pi perform real-time object detection. It's a fun project and I hope you enjoy. Leave a comment if you have any questions or future video requests.
I may earn commission if you purchase from the links below:
RASPBERRY PI 4: amzn.to/3BC4i9I
ELECTRONICS COMPONENT STARTER KIT: amzn.to/3qB23xg
USB 1080P WEBCAM: amzn.to/3Lq0OL7
0:00 Intro
Tensorflow github repo: github.com/tensorflow/example...
Subscribe for more Raspberry Pi tutorials :) ruclips.net/user/samwestbytech...
0:43 Setup Raspberry Pi
Here's my setup tutorial: • Raspberry Pi Headless ...
ssh into your Pi or open a terminal on your pi
Update
sudo apt-get update
sudo apt-get upgrade -y
python3 -V
sudo -H python3 -m pip install virtualenv
mkdir project
cd project
python3 -m virtualenv env
source env/bin/activate
4:00 Install libraries
python3 -m pip install "picamera[array]"
python3 -m pip install tflite-runtime
python3
import tflite_runtime
tflite_runtime.__version__
Both of the commands above should execute without errors
quit()
5:25 Install the example code
git clone github.com/tensorflow/examples --depth 1
sh setup.sh
sudo apt-get install libatlas-base-dev
NOTE:
If the "git clone" command gives you trouble, you can manually download the examples folder here github.com/tensorflow/example...
6:50 Testing!
python3 classify.py - Наука
Thank you thank you thank you so much sir... 🙏🙏🌹🌹
I was doing this object detection using raspberry Pi for my final year project. Because of you I have learned a new thing and saved my thousands of rupees... Again thank you sir... 🙏🌹
great video!
thank you for teaching me how to implement object detection with my raspiberry 4.
NICE!!! Thank you for this video!!!
At least we know that the raspberry has not become self-aware 😅
Great video, thanks for making this!
Awesome video!Thanks so much
Thank you for you brief and precise tutorial ! LMS (France)
really nice, it worked perfeclty with my raspberry pi3 model A. I was identified as a person an my face as a dog which is quite accurate.
Very cool, is there a way to add a highlighted box around the object?
amazing, worked :)))))))))))))))))))))))))))))))) thanks alot
I got to the end and it asked to install open cv2, so do I need to install that into the env?
can you please make a tutorial on training your own data using google colab?
Hi sir your tutorial was the one that worked after 3 days of finding. I wanted to know how to shutdown the video feed? And how can i get the boundries around the object? And how can we make it a standalone live feed system to a android phone or tablet? I mean i dont have to put the commands all the time. Thank you.
hey this is so cool ,
i have a question ... what if i take your code(python) and trying to connect it with oher devices using C programming ?
Thank you so much for the video, pretty clear! However when I get to the final step after running python3 classify.pi I get errono 2 no such file in directory. Could anyone help me solving this?
I would like to perform object detection on a pi, and don't know if OpenCV or TensorFlow is the best tool. I want train the sw to detect any given object, and I also need an output that indicates the position, direction and (hopefully) speed of a moving object - a "follow me" or "follow a given object" feature. Any ideas or help would e appreciated. I'm a relative novice.
Which algorithm is used for image classification in this tutorial?
Hi Sam, fantastic video tutorial. Thank You. Have you tried (and what is necessary) to run with the Coral USB Accelerator?
I have not but they look really cool!
it keeps saying no module named tflite_support is found. can someone please help me?
Thank you so much bro. can you implement Real-Time face emotion detection using raspberry pi 4 because I have a task about it and I got some error can you help us to implement it. Thank you again.
Have a nice day and coding.
I get an error in the end. Last command (python3 classify.py) starts but then stops and says "struct.error: unpack_from requires a buffer of at least 4 bytes". How to fix this?
Hi, I get following error when I try to install tflite-runtime using command "python3 -m pip install tflite-runtime"
ERROR: Could not find a version that satisfies the requirement tflite_runtime(from version: none)
ERROR: No matching distribution found for tflite_runtime
do you have a guide in running yolov4 tiny on raspi
Hi, i noticed that the video streaming of your Pi have 14 fps on it. I have tried installing Ultralytics YoloV8 on my Pi with the models also exported to NCNN. Now I also use a 720p webcam, however, my Pi can only give me 2fps max. I wonder if you were using a Google Coral USB Accelerator making it possible to reach 14fps for your stream. Did you? Or was it just the Raspberry Pi's power to deliver 14fps. Or does the brand of webcam even matters when it comes to fps? Would greatly appreciate a reply.
Thanks.
PS. I am using a Raspberry Pi 4B 8gb
Hi, thank you for that great video. Do you think it’s possible to take a Foto, select an area in it and then compare the next photos to that area and check if the same object is in it?
Looking for a similar solution. Did you find anything?
Unfortunately not. I solved my problem in a different way. wanted to check if tax stamps are applied to a product. I just got a luminescence sensor now. @@MidwestMusicEnthusiast
Couldn't get it to load numpy for pi camera array . That pretty much stopped me right there is there some numpy package not mentioned here that must be installed first?.......it went on the third try wonder why ahhhhh the wonder years 😁
Thanks for this video Sam. I ran a similar set but my version has bounding boxes around the objects. It also always detects my hand, bounds it and labels it “person “.
I want to do object detection of a custom data set, specifically I want to use the pi to identify tree species from their bark. Do you know how to make a custom labelled data set to be used as training data and how to implement it?
Did you ever get a reply about this?
@@shahriarrashid1071 no reply. It’s for a system to detect Ontario tree species from just the bark pattern. I’ve been assembling image sets for training, I understand I need a lot of images but am not sure how much is enough.
@@charlesblithfield6182 i tried to detect plant disease from leaf images and i used about 500-800 images per disease for a test accuracy of 90%+.. so i guess same could apply to you? (i used a version of the plant village dataset available online)
how is the fps on this one soo decent, i did couple of projects on opencv and all of them were really slow despite few optimizations
Can I stream this opencv object detection footage to a webpage?
Hi Sam ,When I try to import tensorflow I get this error, Descriptor can't be created directly
Hi sam can you help me? please, so sorry for my bad english but i have an error when i copy and paste in raspberry, the raspberry asks me for a user and password from git hub and i dont know what is that
After about 19 failed attempts, this one finally worked...
My issue was trying to use Bullseye - PiCam and VNC... a combo that doesnt work!
How did you fix this ? I'm running into the same issue
Thank you for this nice video.
But i have just a IP-camera with a rtsp-stream. How can I use this cam?
I haven't worked with wireless cameras before, sorry
I need help i tap this sudo apt-get upgrade -y and you wait like 5min me i juste have this line apt-listchanges : Reading changelogs. . . for like 45 min do you know why ?
Raise 0SError (port audio library not found) how do I fix this
Hi Sam, I get a ModuleNotFoundError: No module named 'tflite_runtime' in the python process, what should I do?
I have followed the process well
Sir Help me on how to draw bounding box on detecting picture
Damn, bro. I am more impressed with your jawline😂
hey how do i do this with pi camera 3
hey very good till i got to the last line of code with this error
Traceback (most recent call last):
File "/home/pi/project/examples/lite/examples/image_classification/raspberry_pi/classify.py", line 20, in
import cv2
File "/home/pi/project/env/lib/python3.9/site-packages/cv2/__init__.py", line 5
from .cv2 import
^
SyntaxError: invalid syntax
how do Ifix?
I have this same issue, by anychance were you able to fix?
Hey its a good work, can we add another webcam to detect the object that detecting object using multiple cameras?
Yes that should be possible. You'll have to run two scripts at the same time accessing the separate cameras
What should I write to run by usb camera instead of picamera[array] ?
Can I use this tutorial with python3.7.3 version ?
Can you use a webcam with the raspberry PI and this program as well.
Yes you can. That is what I'm doing
I've got an error missing cv2. Does this needs installing Open CV before anything else?
Where you able to fix it? Same error
would it be possible to differ between different cats (they have differnt fur pattern)?
do you already know? because im also want to different 2 different cats, longhair and shorthair
Will it work on raspberry pi 2?
Mini robot based to get and place objects using tf
Sir can you make a video “how to use Resnet50 in raspberry pi for face recognition “ thank u
Hai Sam great staff going on here I have been trying to implement a new one security camera that detects an object and unlike sending email but it gives you a call.. when object is detected..
Sounds like an interesting project!
@@SamWestbyTech yeah and I thought you could help
Great Video, thank you so much...
do you have a camera pi setup video?
Thanks! No, sorry I don't have a pi camera setup video
Hello, i followed your guide, but i am getting this error code: ImportError: libIlmImf-2_5.so.25: cannot open shared object file: No such file or directory, i cant seem to get around it, any ideas how?
run this sudo apt-get install libatlas-base-dev
what web cam do u use for the project. or what would u consider
I have the logitech 720p. It's easy and cheap
awesome video!
but everytime after I trued to download the example I am asked for an username and password. I there something wrong on my end?
Ahh thanks for pointing this out. This likely means you don't have git setup. You don't have to set it up. You can download the files manually here github.com/tensorflow/examples/archive/refs/heads/master.zip
Just download and extract the zip and put it in your project folder
Hi I'm trying this on a Raspberry pi4B with Debian 12 and raspberry pi NoIR camera V2 and I'm getting and error: "Unable to read from Webcam. Please verify your webcam settings." Could you please advise.
got the same issue, if anyone knows how to resolve it, I would welcome some help
did you figure it out? have the same problem
@@tammymathews8455 i downgraded to debian 11 bullseye to make it work.
Can I get the list of objects this code can detect using the pi board and camera?
could you help if you did
hi, i need help with the script
Hi I get an error everytime I try to install picamera array. I used a fresh SD card with Raspberry Pi buster legacy.
note: This error originates from a subprocess, and is likely not a problem with pip.
ERROR: Failed building wheel for numpy
Failed to build numpy
ERROR: Could not build wheels for numpy, which is required to install pyproject.toml-based projects
I have this same issue
did you find the solution to it? i'm cramming a school project and ran into this problem @@mxchaotic
Try python 3.7
Failed to import tflite_runtime: cannot import name 'core' from 'tflite_support'
hi bro now 2023 , raspberry pi source/apt not upgrade (connection error)
I want to make when any object detect then voice output via speaker so know which object detect
Yes you can! You can edit the sample python code take the object it detects and then use a text to speech library to read that through your speakers
@@SamWestbyTech I want to glow led when my custom object detect.where I put this code for this?
@@SamWestbyTech thanks
Hello, Sam. I am trying to run this script with a Google Teachable Machine model, but I got some errors. Do you know if it's possible to use this type of model?
If you could make a video about it that would be great.
What kind of errors do you have? I found this tutorial on Teachable Machine and TFLite github.com/christianversloot/machine-learning-articles/blob/main/using-teachable-machine-for-creating-tensorflow-models.md
OMG Teachable machine looks great to create models.. the easiest way possible, thank you
were u able to figure it out? I'm also trying to do this. Need it urgently for my final year project
I have picamera but your code can not run with me.
look what is the error unable to read from webcam
The same problem for me also... it shows an error.
"ERROR: Unable to read from webcam. Please verify your webcam settings."
could you find a solution
I used the object detection for my graduation project
It’s work at the first time
But when I run it again it’s give me that it doesn’t have a python project file
I returned all the steps and re-download python and I still have the same problem
Please help
Perhaps you need to re-activate your virtual environment
Hi does it work with color classification?
Hi! I wanna know if you made this project?
what is the meaning of
PEP 517
How to enable the bounding box for an object ?
how can i add a bounding box
I am getting error python -m pip install subprocess-exited-with-error
while sh setup.sh pls respond as quick as possible.love your videos bye!
Thanks,
Jaidev
Did anyone encounter the following error? ERROR: Unable to read from webcam. Please verify your webcam settings.
could you find
Hello sir, Many thanks for the video, while running the example fielERROR: Could not build wheels for numpy, which is required to install pyproject.toml-based projects
[end of output]
note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-errorpip install numpy
any idea anyone has how i can remove from this error
I am using python 3.11.2 and tpu 2.14.0
were u able to solve the problem?
Hey I am getting error message of : externally -managed-environment
in setup of virtual environment
do python3 -m venv venv then source venv/bin/activate and type the command again
When I run the code classify.py, it shows an error.
"ERROR: Unable to read from webcam. Please verify your webcam settings."
can anyone help me
@@mathusudhanbaskaran8497 Try this:
sudo raspi-config
Select 'Interface Options' and then enable 'Legacy camera support'. Finish then restart your Raspberry Pi and then run classify.py again.
@@Andreas2048 didnt work can you help me
How do I activate it if I use rpi camera?
It says no module named cv2
Hii sam, its been 1hr but sh setup.sh is not executed completely, it just paused, I tried many times but always stopping at the same line(i.e; Installing build dependencies).
Please help me to resolve this problem .
I had an issue when using pyton 3.11, try using 3.7
were u able to solve the problem?
import cv2 no module named cv 2
please help
Make sure you reactivate your virtual environment
With that thumbnail you make it seem like this is real object detection with positioning information, while in reality this is only image classification. I don't think that you're misleading people on purpose, and many viewers seem to like the solution, but it would save future viewers time when they are looking specifically for object detection ...
is there a way to make a custom object detection and or after training your own model and adding that to make it classify objects that would be much appreciated if you could contact me as I need this for dissertation work
It is possible. Look up a tutorial for “transfer learning object detection”