293 - Denoising RGB images using deep learning (Noise2Void)

Code generated in the video can be downloaded from here: https://github.com/bnsreenu/python_fo... All other code: https://github.com/bnsreenu/python_fo... This video covers the topic training a Noise2Void denoising model using RGB images. Images need to be in lossless format (e.g., png or tiff). JPG format is not allowed if you're using the DataGenerator object but you can use skimage or other libraries to load them. The same approach can be used for grey-scale images, for example SEM or CT images. Next video in this series will cover multichannel and 3D image denoising. Noise2Void: https://arxiv.org/abs/1811.10980 https://github.com/juglab/n2v Noise2Void learns directly from noisy images without the need for clean images, making it the ideal choice for denoising confocal images.​ Only requires noisy images, no need for additional noisy or clean images. ​ Assumption:​ Signal has a structure and noise does not. ​ Therefore, it is possible to predict signal by looking at the surrounding pixels but impossible to predict noise. ​