Putting in TensorFlow Lite on the Raspberry Pi

On this information, we will likely be displaying you how you can set up TensorFlow Lite on the Raspberry Pi.

Raspberry Pi TensorFlow

TensorFlow is an open-source framework developed by Google for machine studying and synthetic intelligence. You need to use this for numerous duties similar to classifying a picture, detecting the bounding field of objects in a picture, and even estimating the pose of individuals.

TensorFlow Lite is a light-weight model of “TensorFlow” designed for low-powered gadgets such because the Raspberry Pi.

You cannot use the Lite model of TensorFlow to coach fashions. You'll be able to solely use it to run pre-trained fashions which have been made suitable with the “Lite” model.

Fashions designed for TensorFlow Lite should be light-weight and fewer computationally costly.

To indicate you ways this all works, we may also present you how you can use the instance “Picture Classification” mannequin together with your Pi digital camera or webcam.


Under is the gear we used when putting in the TensorFlow Lite software program onto the Raspberry Pi.


Raspberry Pi

Micro SD Card

Power Supply

Ethernet Cord or Wi-Fi

Raspberry Pi Camera or USB Webcam

Optionally available


USB Keyboard

USB Mouse

Raspberry Pi Case

This tutorial was examined on a Raspberry Pi 400 operating the desktop model of Raspberry Pi OS Bullseye.

Getting ready your Raspberry Pi for TensorFlow

Earlier than you'll be able to set up TensorFlow, we have to full some preparation work. Sadly, TensorFlow Lite isn’t out there by the included repositories. As a substitute, we should depend on Google’s package deal repository.

1. Our first step is to carry out an replace of our Raspberry Pi’s package deal record and improve any current package deal in your system.

To carry out each of those updates, you'll need to run the next two instructions inside the terminal.

2. As soon as the replace completes, we might want to add the Google package deal repository containing TensorFlow Lite to our Raspberry Pi.

We are able to begin this course of by including the repository to our sources record through the use of the next command.

3. Though we've added the repository, we nonetheless want so as to add its GPG key into our keychains listing.

The package deal supervisor will use this key to assist be certain that the file did, in actual fact, come from this repository.

Obtain and save the GPG key to our keyrings listing through the use of the next command in your Raspberry Pi.

4. Since we modified our Raspberry Pi’s package deal sources, we have to replace our package deal record to scan the newly added repository.

Carry out an replace of the package deal record through the use of the command beneath.

Putting in TensorFlow Lite to your Raspberry Pi

Now that we've ready the Raspberry Pi, we are able to set up the TensorFlow Lite runtime to our Raspberry Pi.

1. To put in Tensorflow Lite, all you must do is run the command beneath in your gadget.

This may set up the newest TensorFlow Lite runtime from Google’s package deal repository.

2. Now that we've put in the package deal, we are able to confirm that TensorFlow Lite is now working by importing it.

You can begin the Python command-line interface (CLI) in your Raspberry Pi by typing within the command beneath.

3. Throughout the Python CLI, it's simple to confirm that TensorFlow Lite is put in.

All we have to do is use the next line inside the interface. All this line is doing is importing the interpreter library.

If every part has labored up to now, it's best to see no additional messages inside the command line. Now you can run your TensorFlow Lite fashions in your Raspberry Pi.

Operating a TensorFlow Lite Mannequin on the Raspberry Pi

There are numerous pre-trained TensorFlow Lite instance fashions on the official TensorFlow website.

You will discover examples with guides for the Raspberry Pi by on the lookout for the “Strive it on Raspberry Pi” textual content.

We will likely be utilizing the “Picture Classification” mannequin for this instance. First, be sure to have a digital camera related to your Raspberry Pi. This digital camera can both the the Pi Camera or a USB webcam.

1. To start out this off, we will likely be cloning the examples instantly from the TensorFlow GitHub.

Nonetheless, to clone the software program, we might want to set up the “git” software program to our Raspberry Pi.

2. With “git” put in, clone the instance repository utilizing the next command.

Utilizing the “--depth 1” possibility ensures we don’t clone some other repositories referenced by the one we're cloning.

3. We are able to now grow to be the “image_classification” instance listing.

That is the place the Python script sits in addition to a setup script that can obtain the mannequin that we want.

4. To run this script, we have to modify the file’s permissions to present us execute privileges.

Use the chmod command beneath in your Raspberry Pi’s terminal.

5. We are able to run the included setup script with every part now prepared.

Utilizing this script will set up any dependencies required by Python and obtain the pre-trained TensorFlow mannequin.

6. Now, run the picture classifier utilizing the next command in your Raspberry Pi.

Upon operating the script, it is going to open a window in your gadget. You will note a video feed out of your digital camera displayed right here.

Within the top-left of the window, you will notice some textual content. It provides three guesses on what that picture may very well be. Every guess has a likelihood related to it. All likelihood values will add as much as “1” in whole.

If you happen to see a fair unfold throughout these values, it signifies that the mannequin can’t confidently acknowledge the picture.

python3 classify.py


At this level, it's best to now have efficiently put in TensorFlow Lite to your Raspberry Pi.

Throughout this tutorial, additionally, you will have had an opportunity to check a pre-trained picture classification mannequin.

If in case you have run into any points with getting TensorFlow Lite put in and operating in your gadget.

Be sure you try a few of our IoT projects or a few of our great Raspberry Pi projects.

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *