Your first anomaly detection model in 5 minutes!

The most important part of any AI solution is the data. That’s why our ‘Image Explorer’ is at the heart of the platform.

Upload your image data in seconds and manage your image data centrally. This means they can be shared with colleagues and accessed from anywhere.

Select all ‘normal’ image data with the label ‘normal’, all error images with ‘anomalous’.

Thanks to our in-built data versioning, you can always return to a previous state.

With our 3-step quickstart tool, you can train your first anomaly AI model in under 5 minutes.

Select a data set and an AI model. You can choose between the best public models from research and models developed by preML.

The image data marked with the ‘normal’ label is used for training. ‘Anomalous’ and some “normal” images are retained in order to evaluate the quality of the model. Depending on the size of the data set, the training runs for a few seconds to minutes.

If you are not yet satisfied with the quality of the output, you can use our analysis tools to go deeper and make adjustments.

A useful analysis tool is the threshold analysis plot that can be created for each training session. Here you can easily define visually how sensitively the model should react to errors.

You can also visualise the output for each image in the dataset, allowing you to correct your training data to improve the model.

Would you like to test your trained model in practice? No problem! Simply connect a USB webcam and check your components.

You can also try out your first system without any expensive hardware. Any commercially available USB camera can be integrated as a webcam. You can also use this to create your data set. Have fun trying it out!