Tensorflow Lite with Object Detection on Raspberry Pi!

Hey welcome back, Ben again! Today we are looking at how to install and use Tensorflow Lite (tflite) on a raspberry pi and do some simple object detection! This setup is surprisingly straightforward and pretty quick to do!
Make sure to use the links below, especially if you run into the GLIBCXX_3.4.29 issue!
Links and resources:
Virtual environments: www.freecodecamp.org/news/how...
Tflite quickstart guide: www.tensorflow.org/lite/guide...
Object detection repo: github.com/tensorflow/example...
GLIBCXX fix: stackoverflow.com/questions/7...

[@pocketgamer8446]

Thanks A LOT, I've been struggling to setup this for days and your simple tutorial helped me ! :D

[@elvenkim]

Like your enthusiasm! Keep it up!

[@tanmayhannikeri4130]

thanks a lot bro this vidio helped me a lot and I'm really grateful for it. keep doing what your doing.and all the very best in your youtube career ;)

[@Jerseylance1]

Nice work, but what if you want to use pan&tilt servos to have them follow you?

[@VinhTran-mq6iz]

Thanks a lot

[@valixg7557]

Thank you very much. I tried to Install Tensorflow Lite myself, but failed. Well I had luck that I found your Video. If you have time, can you make a Video were you use a speech AI to regognice the person and when you give him a command, that he do it. Like write "Blabla" or " move", shutdown, etc. Well it would help me in my Projects.😅😅 Thank you very much that you invest so much time doing this things.😊

[@maysalsayed2343]

ok , what if i want to run ssd mobilenet v2 on raspberry pi? i've watched your video of it but i need to apply it on my raspberry pi

[@armenkalaidjian4494]

Doesn't work on Debian version: 12 (bookworm). Unable to install tflite_support

[@nikitagromov9717]

Sorry, but which RPi do you use? Is this a Raspberry Pi 4 ?

[@nanaasanteful]

My students have been trying on a number of videos to use a custom model and it seems there's an error with the tflite generated from colab. "RuntimeError: Input tensor has type kTfLiteFloat32: it requires specifying NormalizationOpt ions metadata to preprocess input images." is there a specific version of tflite to generate or what could it be?

[@fajarrevo5200]

how to change image or upload image for detect?

[@nitinmohariya1002]

Hey mate I just want the code to detect person can you suggest me how to modify code for that?

[@ahmedakyol8830]

Hello thanks for information but i have an eror after "sh setup.sh" my setup is raspberry pi 5 and picamera3 how can i fix it

[@user-ub1gn6zh9z]

"RuntimeError: Unable to open file at tree.tflite" ("tree.tflite" is my own custom model") Can you help me?

[@vasylcf]

It was interesting, thanks

[@maysalsayed2343]

how can i let it work for pi camera?

[@akshespage5172]

hello sir all the step I did correct at the last step's before is i got the error ''import error: numpy.core.multiarray failed to import'

[@mohamaddabdoub9527]

Hi, Could you please do a face recognition tutorial that can detect general faces not just like mine? Thank you very much

[@supriobarua53]

just a suggestion, while making the video, please fit your interface wherever you are working with the screen, or maybe zoom a bit. Its really difficult to see what you are typing and all for the people who have eye problems like me :D

[@GoyavoCity]

Cool stuff

[@yoshithaab7290]

Is 'camera'/'legacy camera' option available under 'interfaces' tab, in raspberry pi 4b+, because our camera is working fine, but we are not able to see that option in 'interfaces' tab, could you please help!

[@ManojReddy-xv1sy]

I am also having same problem have you completed . If completed please help

[@yoshithaab7290]

@@ManojReddy-xv1sy I'm still facing the same issue

[@aymeneboucha6057]

you don't need it in the latest version, the camera is activated by default, there is no need to go activate it from interfaces.

[@yoshithaab7290]

@@aymeneboucha6057 But, for some projects like object detection, they say we'll have go to the interfaces tab, select that specific port related to pi camera and proceed with other steps...so how to do the same in raspberry pi 4 b, as it doesn't have that option...could you please help!

[@aymeneboucha6057]

@yoshithaab7290 yes the camera is enabled by default, if you want the option of legacy camera you need to switch to the bullseye OS ( 32 bits legacy ), and by the way apparently OpenCV is not compatible with the new OS only with 32 bits legacy

[@aminmutalib874]

i got problem: 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 may i asking for some help here?

[@maysalsayed2343]

i downgraded the python version for the virtual environment to python 3.7.3, and it worked

[@dgeorgakopoulos]

I am using a Raspi 3 A+ and a camera module 3 NOIR, at the last step I get an error saying unable to read from webcam. Please verify your webcam settings. Can anyone help me here?

[@k.pradeepkumar968]

@dgeorgakopoulos ,me too facing the same issue ,

[@maysalsayed2343]

did you solve it?can you please tell me how😊

[@ViciOW2]

Hey I've been following the tutorial and when I run "sh setup.sh" like at 5:30 in the video I receive an error: ERROR: Failed to build wheel for numpy Failed to build numpy If I then try to run "python3 detect.py --model efficientdet_lite0.tflite", I receive: ModuleNotFoundError: No module named 'cv2' Any help would be appreciated.

[@s.rmishra7176]

im getting the same error the cv2 error gets resolve₫ when i download opencv seperatly but theirs another error and like no tflite-support task found

[@s.rmishra7176]

got the solution man just downgrade the python to 3.7 and it works

[@aymeneboucha6057]

@@s.rmishra7176 which version exactly ? 3.7.2, 3.7.5.. ?

[@michalekczer]

Which version of RPi OS do you have? Buster 32 or 64 bit?

[@LazyTech825]

Raspbian 64 bit

[@kurtvdpoel]

@@LazyTech825 Do you mean Bullseye? Could you link to the image you used?

[@arbelcasquejo2336]

Hi, will this work on python version 3.11?

[@LazyTech825]

I don’t see anything about python 3.11 specifically on the website for tflite, but you can try it!

[@shayanajmal4752]

Hi there, I have a quick question. I followed everything step by step and it worked really well until the last step. I get the following error when I run the file: INFO: Created TensorFlow Lite XNNPACK delegate for CPU. [ WARN:0] global /tmp/pip-req-build-hwenbn4e/opencv/modules/videoio/src/cap_v4l.cpp (1001) tryIoctl VIDEOIO(V4L2:/dev/video0): select() timeout. ERROR: Unable to read from webcam. Please verify your webcam settings. I am not sure how to fix this, I am using a raspberrypi4 4gb with the deebian bookworm OS. I downgraded my python version in the examples directory as well to 3.7.3 to match your video. Any help would be greatly appreciated!

[@ManojReddy-xv1sy]

Same for me have you solved it

[@maysalsayed2343]

if you solve it , can you tell us how?

[@2FastEdits8]

I keep getting an error when trying to run the sh setup.sh command. Saying that there is an error from a subprocess. Could you help me. Thank You.

[@aymeneboucha6057]

did you fix it ?

[@Cyonko]

@lazyTech Hi thank you very much, pls help I have raspberry pi camera 3. I install everything, but how I can use my camera, thank you.

[@maysalsayed2343]

did you solve it? can you please tell me how😊

[@revolution322]

Thank you for the tutorial! Does anyone faced error "No module named cv2 "?

[@hanikurdi98]

me to

[@LazyTech825]

Do you get that when installing, or when running the code?

[@hanikurdi98]

@@LazyTech825 when i running the code i already installed the opencv and tensorflow

[@revolution322]

When running the code using the TF virtual environment. If I run a normal OpenCV python code my libraries are ok. Thx! @@LazyTech825

[@s.rmishra7176]

running the code @@LazyTech825

[@benjaminplaczek4763]

Wake up babe new Lazy Tech video dropped

[@LazyTech825]

Hello darling

[@Bengemon825]

Whoop whoop

[@sejongjeon4143]

Hi, i finally input that code python3 detect.py --model efficientdet_lite0.tflite then INFO: Created TensorFlow Lite XNNPACK delegate for CPU. ERROR: Unable to read from webcam. Please verify your webcam settings. i use raspberry pi camera rev 1.2 do i need to modify the code of detect.py ??

[@s.rmishra7176]

did you get any solutions

[@s.rmishra7176]

got the solution it doesnt work for picamera only works for usb webcams

[@k.pradeepkumar968]

Awesome tutorial ,wonderful explanation..... i am using the raspberry pi 1.3 camera (5MP) instead of the usb web camera and getting the following error any one please help me to resolve the issue , python3 detect.py --model efficientdet_lite0.tflite INFO: Created TensorFlow Lite XNNPACK delegate for CPU. ERROR: Unable to read from webcam. Please verify your webcam settings.

[@maysalsayed2343]

did you solve it? can you please tell me how😊

[@respectvideos1929]

bro how can install tensorflow on my raspberrypi

[@LazyTech825]

That’s what this video is about??

[@pocketgamer8446]

If someone encounter this error : INFO: Created TensorFlow Lite XNNPACK delegate for CPU. qt.qpa.xcb: could not connect to display qt.qpa.plugin: Could not load the Qt platform plugin "xcb" in "/home/raspberry/examples/env/lib/python3.9/site-packages/cv2/qt/plugins" even though it was found. This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem. Available platform plugins are: xcb. Aborted Just enter this command and it'll be fixed, it worked for me : export DISPLAY=:0.0