LumiScan Training

Download this manual for offline use (ZIP, 5.4 MB - HTML and Markdown, version 2.3.0.3)
Three folders of annotated images and a JSON configuration

Prepare the data. Annotated images for training, validation and test.
Training curve falling over the epochs

Train the network. From scratch or continuing an existing network.
ONNX and HEF model files for AI Object Matching, LumiScan VGR and Hailo TPU

Export the model. As ONNX, or converted for a Hailo TPU.

LumiScan Training trains neural networks for object detection and localization, from scratch or by continuing an existing network. Training, validation, testing and inference are set up in one JSON configuration, and the finished model is exported as ONNX for LumiScan AI Object Matching and LumiScan VGR. The optional Hailo PlugIn converts it for a Hailo TPU. It runs on Windows 11 with an NVIDIA RTX graphics card, locally or on a shared training server.

For licensing, see Licence. Questions or an error message? See the FAQ.

In this manual

  • Description – LumiScan Training trains neural networks for object detection and localization, controlled via JSON, with ONNX export and optional Hailo model conversion.

  • Requirements – Requirements for LumiScan Training - Windows 11 Professional or Enterprise, NVIDIA RTX graphics card, CUDA 11.8, CodeMeter 9.00 and the Hailo PlugIn.

  • LumiScan Training Installation – Installing LumiScan Training - run the setup program with its own Python environment and, for a Hailo TPU, optionally install Podman and the Hailo PlugIn.

  • User Interface – User interface of LumiScan Training - elements of the standard interface after program start and why the JSON settings must be configured before training.

  • App Settings: Local and Remote Backends – App settings in LumiScan Training - use the local backend or set up a shared backend server and connect to it from a web browser or another installation.

  • JSON-Settings – JSON settings of LumiScan Training - example configuration and parameters for training and validation data, classes, network type and export.

  • Model Conversion: ONNX to Hailo – Model conversion in LumiScan Training - convert a network exported as ONNX into a Hailo model in the Hailo conversion container, using the Hailo PlugIn.

  • FAQ – FAQ on LumiScan Training - requirements and installation, Hailo PlugIn and Podman, JSON configuration, training, model conversion, licensing.

  • Troubleshooting – Problems with LumiScan Training - model conversion, GPU memory, shared backend and licences - causes, checks, solutions and the information support needs.