We have many reasons for building this platform. One of them is that we want to make AI available, safe and useful to end users. To benefit from AI today, you need a large team of engineers and if using standard methods, the built AI models will not a have any sense of uncertainty. We call this the problem of uncertainty and we think using such AI models are outright dangerous. The approach we are aiming to take with Alvíss AI is to use Bayesian deep learning, an alternative to conventional machine learning where uncertainty is a first-class citizen.
We want to abstract away the need for running dedicated machines, handling and annotation your data. We also want to offer out-of-the-box working models.
During the lifetime of Desupervised we have built various AI models, ranging from computer vision to forecasting. We publish our most battle-tested AI models as capabilities on our platform, which all started as bespoke projects. Some of the capabilities we offer on our platform are shown below.
Image classification is the task of categorizing events captured in images. This is typically wanted when there's a theme in the image, e.g., detecting a malfunction of a machine or a product during operation. It can be used for
- Diagnosing skin cancer from images of skin
- Detecting and/or predicting an accident
- Lie detection by facial expressions
Object detection helps identify and locate multiple objects in a scene as captured by an image as well as a video stream. It can be used for:
- Parking space availability
- Localization of known objects in an image
- Crowd count in spaces
OCR locate and read texts on an image. It can be used for all kind of tasks that involves reading e.g. scanning receipt, car license plate, package labels, etc.
Time series allows processing tabular data. It can be used for
- Housing pricing forecasting
- Stock prediction
- User survey prediction
Have a look at some of the tutorials for examples and use cases!