Amazon announced a new tool called Amazon Rekognition Custom Labels. This allows machine learning to be used to let AI models recognize certain objects, even if only limited data is available.

Normally, huge datasets have to be used to train models to then, for example, let them recognise a dog in a picture. Amazon Rekognition Custom Labels can train models for very specific use cases without the need for large datasets like these.

“Instead of having to train a model from scratch, which requires specialized machine learning expertise and millions of high-quality labeled images, customers can now use Amazon Rekognition Custom Labels to achieve state-of-the-art performance for their unique image analysis needs,” the company wrote in a blog post.

No machine learning expertise is needed to use the tool, as users are helped through all required steps. Once all steps have been completed, users can get an overview of the accuracy of their model and can request suggestions for possible improvements.

Specific use cases

The company gives the recognition of engine parts in a car garage as an example. Rekognition Custom Labels can recognise which parts they should or should not order extra. This is a task where only recognising the specific car components is important, and thus a lot of time can be saved.

Other users of the tool include NFL Media, a part of the National Football League in the US. NFL Media manages a huge amount of footage and uses the tool to search for certain specific items (e.g. team logos, etc.).