These are our customers most frequently asked questions (FAQ).
In short, the Peltarion Platform is a cloud-based operational AI platform that allows you to build and deploy your own deep learning models, even if you’re not an AI superstar.
To get started, check out our tutorials.
To use Peltarion Platform you basically need Google Chrome installed on your computer. Then you’re good to go! For more detailed requirements, check our details in the Technical requirements article.
If you have any questions about the Peltarion platform functionality, please contact email@example.com.
First, click + New project in the left panel to create a project. Then click on Open. This will take you to the Datasets view.
To get familiar with the workflow, check out our tutorials that will guide you step by step.
Yes, you have to be consistent and perform the same preprocessing for your train and test dataset.
If you use the Deployment API instead of a downloaded .h5, the same data processes that have been done during training will be performed automatically. That is, you don’t need to do anything.
You can’t use a test dataset on the Peltarion Platform right now. But you can use our tool Sidekick.
But first, make sure that your deployment is Enabled. This is how you can enable a deployment.
Use Sidekick! Sidekick handles the mundane tasks like sending data examples to the deployment endpoints to get predictions.
There are several error messages on the Platform. We try to make them as short and full of information as possible. But if you want further guidance, check out cause and remedy for each error message in Error messages.
Yes, you can have multiple inputs, e.g., in our Tutorial - Predict California house prices, we use both images and tabular data as input.
It depends on your data and on your task. Check out our cheat sheets for tips and tricks on how to solve your problem.
As many as the shape of your target, so it depends on your task.
Binary classification: The last Dense block must have 1 node.
Set Binary crossentropy as target loss function.
Multiclass classification: The last Dense block must have the same amount of nodes as the number of classes. For example, cifar10 has 10 classes to predict, hence your last dense block must have 10 nodes.
Set Categorical crossentropy as target loss function.
Regression: As many as the shape of your target. In the tutorial Predict California house prices the number of nodes in the last Dense block is set to 1 since we want to predict 1 thing, the price.
The target loss function is set to Mean squared error.
This is explained in our webinar: Hands on with deep learning and predictive maintenance.
For more info check out our cheat sheets for tips and tricks on how to solve your problem.
Easy experimenting is a key feature of the Peltarion Platform.
You can run multiple experiments at the same time. If you notice that an experiment is performing better, you can easily duplicate it. Click on the Experiments option menu () and select Duplicate. Name the new experiment and then decide if and from which checkpoint you want to copy the weights.
Continue experimenting in the new experiment and then keep duplicating, tweaking and running until you are happy with the result!
Read more about working with multiple experiments.
First, follow the instructions in the Deployment view part of the Knowledge center.
To reset your password, go to this link: Reset password.
Enter your e-mail and click Send.
You will now receive an e-mail with instructions on how to reset your password.
To upgrade your usage plan, please contact firstname.lastname@example.org.
Stay in the know by signing up for occasional emails with tips, tricks, deep learning insights, product updates, event news and webinar invitations.
We promise not to spam you or share your email with any third party. You can change your preferences at any time. See our privacy policies.
Please check your email inbox account to confirm, set, or update your communication preferences.