Assistance from A to Z

Yes, you can also use images taken with a smartphone for the training in IDS lighthouse, as the import can be done via a simple folder storage.

In our Vision Channel webinar "How to evaluate AI Vision without camera" we show this use case step by step.

Yes, image datasets can be exported via the "Basic Settings" dialog and imported in any user account.

Image datasets can also be "shared" with other users within IDS lighthouse. This enables collaborative work on the same data. To do this, use "Collaboration" in the basic settings of a dataset.

Our hosting partner, Microsoft (Azure Cloud), works in compliance with strict security specifications and with a number of ISO standards, which means that your data is very safe.

It remains entirely your decision who may have access to your data. For additional information on the use of IDS lighthouse with the hosting provider, please see our Terms of Use.

To run neural networks in IDS NXT cameras, an application workflow is needed. A so-called vision app determines how and when images are captured and fed to the AI-based image analysis. The further processing and communication of the results is also determined by the vision app.

For classification and detection there are vision apps available. You only need to configure these with your self-trained networks.

In the cloud-based AI Vision Studio, you can also build your own individual workflows as vision apps for your networks using the block-based editor or the use case assistant. You can also evaluate these with test data before loading and executing the network and app on the camera with the IDS NXT cockpit.

Watch our tutorial video for app creation.

The use case assistant creates simple vision workflows that you can use to test and run your classification, detection, and anomaly networks in an already basic way.

To realise your own individual workflow, you can further develop the assistant app in the block-based editor or completely start there from scratch according to your wishes.

Alternatively, it is also worth taking a look at our online marketplace visionpier. There you will find turnkey solutions, also for IDS NXT.

It is currently not possible to create your own function blocks for use in the editor.

The C++ development environment IDS NXT Vision App Creator offers much more possibilities for the design of vision apps.

Connect to your camera with the IDS NXT cockpit. Then you can load and activate your trained neural network from the AI vision studio with a vision app.

If you have also created a vision app in IDS lighthouse, the procedure is quite similar. In IDS NXT cockpit, use the Vision App Manager to load, configure or delete apps.

Maybe the camera firmware is not compatible with the version you built your vision app against in AI Vision Studio. Updating the camera firmware might help. To the latest camera firmware

You can both check the firmware and update it right away via the IDS NXT cockpit.

If you still have problems with it, our support team will be happy to help you.

You can use the vision apps created in IDS lighthouse on any number of IDS NXT cameras without additional runtime license costs or technical restrictions.

The only costs include the all-in user license of the AI Vision Studio and the cameras.

The ANNs cannot be used with random cameras. IDS lighthouse was specially developed to train neural networks for image processing tasks that can be used with IDS cameras and software. IDS NXT cameras of the rio/rome series are equipped with an AI core developed by IDS, which performs the inference in a hardware-accelerated manner. IDS malibu models work with a special AI accelerator from Ambarella. IDS lighthouse provides the networks for the respective platform in the appropriate format.

Multiple vision apps may be executed simultaneously.

However, only one vision app can be active that uses the AI accelerator of an IDS NXT camera to execute CNNs. If you want to execute several AI-based analysis steps in sequence, this is also possible in a single app. With the block-based editor in IDS lighthouse, you can easily build such sequences together.

Yes, this is possible for IDS NXT cameras of the rio/rome series, but they have to be converted into a suitable format. Neural networks on an IDS NXT camera are hardware accelerated by the AI-Core - thus they must be available in a special format. Neural networks trained with IDS lighthouse are already available in the correct format and can be loaded directly onto the camera. If you already train with Keras (with TensorFlow Backend), the conversion of network architectures such as MobileNet is possible with little effort. IDS provides special tools on request. If you need more detailed information or assistance with the conversion, please contact IDS Sales.