AI Training Service.

Your custom-made automatic point cloud classifier – trained and tested for your use case.

What the AI Training Service achieves.

Manually annotating data in 3D point clouds is a very time-consuming process.

This effort of manual annotation can be significantly reduced with our AI-based training service. It offers you automated and customized classifications with the help of AI. We train a model just for your use case, to give your performance a boost and open up new possibilities.

Our clients reported for some scenarios that the classification effort could be shortened from several days (with manual annotation) to a few hours when using our AI Trainig service.

a graph showing the advantages of the AI Training Service over manual 3d point cloud annotation
I want to try that service

How does the AI Training Service work?

You can either easily generate training data yourself in our Pointly platform or through our labeling workforce. If you have already labelled data, you can leverage this as training material for your classifier and get a jump start with our Service.

Our team of AI experts will accompany you in this process, defining KPIs according to your needs and regularly updating you on performance.

To start the training of your AI model, you’ll need to upload your data to Pointly and manually classify your point clouds to generate training data for the model. As soon as you have a decent amount of training data we can start training your classifier.

a graphic with gear on the subject of AI Training Service

Then the model can take over to classify your point clouds automatically and you only need to make manual corrections. The more often this process is repeated, the better and more robust the model gets. So do the manual work a few times and then rarely ever again.

a graphic with a gear that says that thanks to the AI Training Service the point cloud classification now runs automatically

How you get your hands on the new AI Training Service

  1. Please inquire the service via the form down below.
  2. You receive a Pointly Professional Account to upload your point clouds. The data is split into three batches. Each batch is labeled, either by you or through our labeling workforce*. We run automatically optimizing trainings for each batch and report the performance back to you.

Do you want to know more or do you have any questions? Please read our FAQ or send us an email to!

*Labeling fees apply at competitive rates.

What you get:

  • A Professional Pointly subscription
  • A personal consultation session with our AI experts to assess your goals and define KPIs
  • Regular performance updates on the training progress
  • An AI point cloud classifier, tried and tested for your use case that you can keep using and training

Please fill out the form down below to inquire for an AI Training Service with us.

Training Service FAQ.

Simply tell us about your use case in the linked form, it takes less than 3 minutes! Or send us an email to

Once we have trained a neural network classifier for your use case, we can use it to classify new data automatically.

  • We will provide you with a Pointly Professional License. That means you can upload your point clouds into a custom project.
  • Your data already contains labels? Great, we can start straight away with training a classifier on those.
  • If you don’t have labeled data, don’t worry: Pointly was built for quickly generating training data. You can either do it yourself or order a labeling service through us.

5-15 days of labeling work provide a realistic start for a Proof of Concept. That means you will get a solid estimate of the future potentials of AI but not yet a production ready model. This number can of course vary a lot depending on your use case and the desired performance. We are offering you a briefing call where we clarify all specifics that apply to your use case. You can choose to either invest that effort yourself or hire a labeling service from us at competitive rates. More training data is naturally always a benefit.

The first and most important step is that you choose a test site to be put aside. We will not train on this test site but will use it to evaluate the final performance of the neural network. During the training process on the separate training data, we will keep you updated on the current performance based on cross-validations within the current training set.

A process that we have tried and tested in previous projects is to split the data into 3 evenly sized sets, beginning with the first, then adding the second and so forth. Why this iterative approach? Because it allows us to evaluate how performance increases with additional training data and it gives you more flexibility and control when generating training data. After every iteration we provide you a small report on the current performance.

Afterwards you can keep using your neural network for classification tasks as well as re-train it with more data based on standard charges.

Pointly developer icon

Developed by Experts.

Geo Data Science Technology
made in Germany

Pointly is a spin-off of Supper & Supper GmbH from Germany. The company is one of the leading full-service data science service providers in the country and develops customized solutions in the areas of machine learning, artificial intelligence, process optimization and digital transformation.