Personal Image Classifier web app

Hi, I found that there is a Personal Image Classifier web app at https://classifier.appinventor.mit.edu./ I wonder if it is an open-source one so I can run it offline and store my own data.

This is from the second AI with App Inventor teacher resource in the official MIT App Inventor website.

Thanks so much

2 Likes

Hi @Tony_Nguyen,

The webapp doesn’t send any data to the server (you can verify this by looking at the Network tab in your browser’s dev tools). Its sole purpose is to serve up the webapp and backing neural network models. All of the training is done in your browser using TensorFlow.js. All data in the session are ephemeral with the exception of exporting the trained model at the end.

The student who built the tool has since graduated, but I’ll reach out to see if he’d be interested in releasing the sources under an open source license.

2 Likes

In order to use a trained Google Teachable Machine2 model from App Inventor, I placed it on my Android local web server. However, with MIT’s Personal Image Classifier, it is not necessary and can be put into the App, making it very easy to use.
By the way, I have a question. We can download a trained model, can we also download and reuse a project (including training images)?
Thank you.

Thank you, I am now able to explain to the one concerned that data is not sent or stored in the server.
Hopefully, the app will be made open-source

I also understand that you should be able to turn off your WiFi once the page has loaded, since everything that’s needed to run the website is loaded at the beginning. This way you can be sure no data is leaving as you aren’t even connected to the internet.

1 Like

Hello, everyone.

I want to add one thing.
I’m trying to create some applications to demonstrate the excellence of this Personal Image Classifier. One of them is an application that informs only photos that the top of the mountain looks good from the image of the live camera that is photographing the mountain. I wrote the description below:

2 Likes

so nice webapp !