Our platform is designed to easily manage large datasets. There is virtually no limit to the amount of files, and our filters make it easy to find exactly what you need.
Files can be placed anywhere. Naturally, we can arrange the storage, but any S3-endpoint or Google Drive will do.
Easily define your labels online and apply them to the relevant sections in your audio fragment. Overlaps are no problem, and we’ll track who worked on each file, allowing cooperative labeling in team, without headache.
Are you having multiple datasets with different ontologies? Or do you want to add extra samples from a public dataset? Our system makes it extremely simple to group different labels, even from different ontologies, together to create a class.
We’ve also included multiple public datasets such as SONYC and AudioSet and are constantly expanding the list.
We’ve got tools built in to assess the balance and overlap of your labels and datasets – as well as a automatic resolutions for these and other potential pitfalls. It’s also possible to configure the spectral settings of the MEL-spectrograms used for your model, so you can balance precision vs computing time to your requirements, or to ensure the models can work with the data you already have on premise.
Many pre-trained models are available to speed up and improve training-runs, and both simple classification as multi-label classification is possible. In the case of simple classification, a background-class can automatically be added.
Easily follow the different stages of training online and watch the results onfold in realtime. Datapreparation starts immediately when a model is configured and submitted for training. Once this preparation is finished a GPU is allocated ASAP and the actual training is started. You can see the model’s performance evolve from iteration to iteration and can easily assess the correct and incorrect classified samples.
Want to continue a training or change a parameter to improve it? No problem! You can stop and resume the training at any time.
Our PAVEsense sensor networks are designed to revolutionize your road maintenance.
Our extensive knowledge on smart sound capturing helps you to monitor your environment in real time.
In need of a more tailored solution? We can transform our knowledge into solutions that meet your needs.
