A machine learning kit that has been designed from the ground up to include everything you need to train machine learning models in Lobe and deploy them to a Raspberry Pi.
Thank you! And you are right, the general idea of Lobe 1.0 is in your message! For Lobe 2.0 we switched a bit as you can see, and the good news is, you can go and download the app at lobe.ai without the need to wait any longer!
The reasoning behind the change and the why we abstracted some of those details you are mentioning was to actually make it even more accessible for people to be able to build machine learning models. We think that this is a paradigm that should be used by everyone, and that’s why stripping down the onion of complexity was really important for us when we started with this project.
Teachable Machine is a good way to start! Lobe tries to give you more ways to, not only continuously work on your model by adding more images, changing the number of labels, and even re-label a bunch of your images while training a custom machine learning model in the background—but also, it gives you the ability to analyze your results in real time, and test your model while giving it feedback, so the loop to make your model better is continuously happening. I love that about Lobe.
That’s exactly the whole thesis behind Lobe, back to the conception phase of the project, this was the main principle we tried to infuse in our tool—how can we make the process of building software more humane—and machine learning is one of the tools to achieve that—with Lobe, we wanted to solve the aspect of building the machine learning, as that wasn’t humane either.
Hey qwerty456127! We’ve actually had users build models like this in the past due to some respiratory tracking they need to get done in order to do their exercises properly, so they can’t concentrate on counting and therefore, an automated system proved useful for them.
On the other comment, yes! The app you are describing sounds really interesting, and it is something that could be build using image classification, you just need the right images and camera setup, though!
I see. Thank you for explaining. The case you describe didn't come into my mind. Nevertheless I doubt counting is important. Some reps less, some reps more, I believe the feeling you always feel when you've done enough. Or you can set a timer - it's not the number that matters, it's the time you spend under the load anyway.
On the other - the way you do an exercise, small details in your posture and the sequence of changes in it - that's what decides if what you do is going to make you more fit/strong, have no effect or just cause pure harm. It's extremely important (at least, in the beginning) to have somebody qualified to watch how you do it and correct you. Many people prefer to train alone though so they need such an app.
Thanks for your suggestions here. We are always looking at ways to improve Lobe, and the feedback loop of how to improve your model is one of the most important ones for us.
You can read our privacy policy, but more than that, you could use any sort of network traffic visualizer to see that we are not lying we are not in the business of selling data, we are in the business of making machine learning accessible to everyone.
Where is the app-specific privacy policy? the link the footer links to some general "Microsoft privacy policy", which covers all kinds of things and if it has anything specific to this app, it's impossible to find.
We are working on adding more project templates in the future, and Lobe is designed upon the idea that machine learning should be made easy—no matter the problem type you are facing.
Thank you very much Aldipower! I am one of the designers from the Lobe team here speaking. This is what we strived for when we designed the website, and to be honest, when we designed the product, too, so I am really happy it came across this way.
In fact, our process for every feature we work on is the same—we start thinking about it with the way users are going to learn about it in mind, that allows us to simplify the way we talk about it and massage the messaging as much as possible, so when we have to talk about it externally, it's so tested that it just comes natural to us, and hopefully to the world.
