top of page
Search

N2V2 denoising model - BioImage.io


Author

CAREamics


Description

N2V2 is a variant of Noise2Void. Noise2Void is a UNet-based self-supervised algorithm that uses blind-spot training to denoise images. In short, in every patches during training, random pixels are selected and their value replaced by a neighboring pixel value. The network is then trained to predict the original pixel value. The algorithm relies on the continuity of the signal (neighboring pixels have similar values) and the pixel-wise independence of the noise (the noise in a pixel is not correlated with the noise in neighboring pixels).

N2V2 introduces blur-pool layers and removed skip connections in the UNet architecture to remove checkboard artefacts, a common artefacts ocurring in Noise2Void.


  • Input channel: 2D images, grayscale

  • Bit depth: 32-bit float

  • Output channels: denoised image


This model was downloaded and converted from BioImage.io, respecting the associated license (see License section below for more information)


Download

By downloading, installing, copying, accessing, or using the software, you agree to the terms of this end user license agreement.


Download model file and test image (ZIP archive)



Requirements

Make sure you have installed Aivia and the required DeepLearning module (according to our Wiki).


Installation and apply instructions

Aivia is required for applying the model file. You can request a demo copy of Aivia here.

  1. Drag-and-drop the model file into the Recipe Console area; or use the 'Load recipe' option in the Recipe Console to load the model file.

  2. Load the test image (or any image of your own) into Aivia.

  3. If your image contains more than one channel, click on the 'Input & Output' section and specify the image channel you wish to apply the model on.

  4. Click 'Start' to apply the model.


License

Copyright 2025 CAREamics authors, BSD-3-Clause. Full license information can be found here.


Acknowledgement

Bioimage.io is supported by AI4Life. AI4Life has received funding from the European Union's Horizon Europe research and innovation program under grant agreement number 101057970.



References

Krull, A., Buchholz, T.O. and Jug, F., 2019. "Noise2Void - Learning denoising from single noisy images". In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (pp. 2129-2137). doi: 10.1109/cvpr.2019.00223


Höck, E., Buchholz, T.O., Brachmann, A., Jug, F. and Freytag, A., 2022. "N2V2 - Fixing Noise2Void checkerboard artifacts with modified sampling strategies and a tweaked network architecture". In European Conference on Computer Vision (pp. 503-518). doi: 10.1007/978-3-031-25069-9_33

 
 
 

Comments

bottom of page