1. You have licensed the AI Toolkit and activated it under Tools > Toolkit Manager.
  2. Docker Desktop is installed together with the denoising container and running.
  1. On the Analysis tab, in the Intellesis Denoising tool, click and select New.
  2. Enter a name for the new model and click . Alternatively, enter the name and press Enter.
  3. A new setting with the name is created and automatically selected as Model.
  4. Click Start Training.
  5. The user interface for training opens.
  6. In the Images and Documents section on the right, click Import Images.
  7. A file browser opens.
  8. Select the image(s) you want to use for training and click Open. Note that RGB images are not supported for denoising.
  9. For images with multiple channels, a dialog opens to select the channel for denoising.
  10. Select the channel you want to use for denoising and click OK.
  11. The image is displayed in the list in the Images and Documents section. Note that all imported images are included in your model.
  12. Select the image from the list.
  13. The image is displayed in the left image container.
  14. Select the Algorithm you want to use and if you want to train a 2D or 3D model. We recommend to first try the N2V algorithm and only select N2V2 if checkerboard artifacts are a significant issue.
  15. The algorithm and dimensions are selected.
  16. In the parameter section, set the Training Steps and adapt the more advanced parameters, if necessary. Without a GPU, you might want to start with less iterations, e.g., 2k.
  17. The parameters are set.
  18. Click Train and Denoise.
  19. Your model is trained based on the settings and the prediction is displayed in the right image container.
  20. If you are satisfied with the result, click Finish.
  21. All changes are saved, and the training window closes.
  1. You have successfully created and trained a model for denoising. You can now use it to denoise your images with the Intellesis Denoising processing function, see Using an Intellesis Denoising Model for Image Processing, or use it during a continuous acquisition (only supports 2D denoising models), see Using Denoising During Continuous Acquisition.