One question we get a lot around here is ‘how is Levity different from Lobe?’
Lobe is a popular machine learning program from Microsoft that allows you to automatically train a custom machine learning model that can be shipped from within your app.
While this sounds similar to Levity, there are a few differences which we are keen to point out! Both work well but have different strengths and ultimately each works best for a different purpose!
Both focus on classification
Lobe's core focus lies only in image classification - it doesn't do object detection yet (though this is coming soon!).
Levity is also a classification tool. However, besides image classification, Levity also offers text and document classification leveraging NLP capabilities.
Both tools offer direct labeling capabilities for your images.
What each tool is best for
Lobe is best suited for images that require your webcam - for example, they offer a very cool dataset builder for action-based images (their example video shows how to build a drinking water vs not drinking water dataset to build a health tracking app).
With Lobe, you could analyze facial expressions, hand gestures or use it to monitor whether your baby is asleep or awake.
Levity is focused on business applications - making your work less mundane in situations where business tasks revolving around images are core to your value generation (e.g. visual quality control, quality of an ad, etc.).
Lobe focuses purely on what's on the image, while Levity can also handle models that you can tune subjectively: let's say you have a specific style of ads you want to go for. You can train a model to identify freelancers' work by your own style guide for creatives
Handling edge cases
Levity has continuous 'human in the loop' (HITL) functionality. It seems that Lobe only allows for human feedback before you export your model and set it live in your workflow. In addition, it seems it only allows for correcting certain biases in the dataset - not for the model performance itself.
Levity has HITL built-in into our processes that allow for continuous learning and improvement. This means that the model is also able to learn when something new or unexpected happens.
Integrations (import and export)
Currently, Lobe allows for uploading files and webcam images for training purposes - but if you want to use it in your processes, you'll have to export it and build your processes manually.
Levity has focused on building integrations with other no-code tools, and also has a workflow builder for direct integrations with your business tools - or just use an API to trigger the model.
Intuitive interface + no-code under the hood
Both tools are very intuitive, but Levity is truly a no-code tool, enabling people at all levels of the organization to integrate their model seamlessly. Lobe however is perfect for researchers who are interested in getting under the hood and modeling as well.
To sum up
Hopefully, this has helped to explain some of the differences between the two platforms.
Both are great to use - it often comes down to what type of process you're looking to automate.
While Lobe is a desktop app - and free to use - it doesn't offer the capabilities for easy integrations with your currently existing processes. If you however want to learn a bit about models - and use them for your research, Lobe is a great fit! For everything else, there's Levity