Comment on page
Retail Image Classification - Nvidia Jetson Nano
Use computer vision and an Nvidia Jetson Nano to classify various fruits and vegetables in a retail setting.
Even with the current limitations of Artificial Intelligence, it is still a very useful tool, and many tasks can be automated with the technology. As more tasks become automated, human resources are freed up, allowing them to spend more time focusing on what really matters to businesses: their customers.
The retail industry is a prime example of an industry that can be automated through the use of Artificial Intelligence, the Internet of Things, and Robotics.
Computer Vision is a very popular field of Artificial Intelligence, with many possible applications. This project is a proof of concept that shows how Computer Vision can be used to create an automated checkout process using the NVIDIA Jetson Nano and the Edge Impulse platform.
Your first step is to create a new project.
Create Edge Impulse project
Enter a project name, select Developer or Enterprise, and click Create new project.
Choose project type
We are going to be creating a computer vision project, so now we need to select Images as the project type.
- Running the Edge Impulse NVIDIA Jetson Nano setup script
- Connecting your device to the Edge Impulse platform
Once the firmware has been installed enter the following command:
If you are already connected to an Edge Impulse project, use the following command:
Follow the instructions to log in to your Edge Impulse account.
Device connected to Edge Impulse
Once complete, head over to the Devices tab of your project and you should see the connected device.
In this example we will use images from the Apples, Bananas, Chillis and Broccoli classes.
Once downloaded, unzip the data and navigate to the Train folder. Then proceed to upload the contents of Train/Apples, Train/Bananas, Train/Chillis and Train/Brocolli. Make sure to select Automatically split between training and testing, and enter the correct label.
Once you have completed these steps, you should be able to see and filter your uploaded dataset.
Now we are going to create our neural network and train our model.
Add processing block
Head to the Create Impulse tab. Next click Add processing block and select Image.
Now click Add learning block and select Transfer Learning (Images).
Now click Save impulse.
Head over to the Image tab and click on the Save parameters button to save the parameters.
If you are not automatically redirected to the Generate features tab, click on the Image tab and then click on Generate features and finally click on the Generate features button.
Your data should be nicely clustered and there should be as little mixing of the classes as possible. You should inspect the clusters and look for any data that is clustered incorrectly. If you find any data out of place, you can relabel or remove it. If you make any changes click Generate features again.
Now we are going to train our model. Click on the Transfer Learning tab then click Auto-balance dataset, Data augmentation and then Start training.
Once training has completed, you will see the results displayed at the bottom of the page. Here we see that we have 96.1% accuracy. Lets test our model and see how it works on our test data.
Head over to the Model testing tab where you will see all of the unseen test data available. Click on the Classify all and sit back as we test our model.
Test model results
You will see the output of the testing in the output window, and once testing is complete you will see the results. In our case we can see that we have achieved 91.46% accuracy on the unseen data.
Before we deploy the software to the NVIDIA Jetson Nano, lets test using the Edge Impulse platform whilst connected to the Jetson Nano. For this to work make sure your device is currently connected and that your webcam is attached.
Live testing: Apple
Live testing: Banana
Live testing: Broccoli
Live testing: Chilli
Use the Live classification feature to record some samples for classification from the webcam connected to the Jetson Nano. Your model should be able to correctly identify the class for each sample. If you are not getting accurate detections, you may need to provide more data samples, or fine tune the training parameters to increase your accuracy. First, you will want to save a Version though.
We can use the versioning feature to save a copy of the existing network. To do so head over to the Versioning tab and click on the Create first version button.
This will create a snapshot of your existing model that we can come back to at any time.
Now we will deploy the software directly to the NVIDIA Jetson Nano. To do this simply head to the terminal on your Jetson Nano, and enter:
This will then download the built model from Edge Impulse and start local inferencing on the Nano. Keep an eye out for a message that gives you a URL to view the results in your browser.
Here we have created a simple but effective solution for classifying various fruits and vegetables using computer vision powered on an NVIDIA Jetson Nano, using Edge Impulse.
You can train a network with your own images, or build off the model and training data provided in this tutorial.