N2V, training and predicting#
N2V is a bit different from traditional deep-learning algorithms. Here are a few more questions to help grasp why. 
Training on a single image
Can you train N2V on a single image?
N2V learning
What does N2V learn?
Incidentally, that is not a philosophy question. 
Predicting on different images
Can you predict images using a N2V network trained on different images?
Multichannel
Should you train a N2V model using different channels (different structures) from the same experiment?
Deconvolution
Is N2V some sort of deconvolution?
Pre-processing
Here are two cases of pre-processing:
My images come from the same experiment, but there was drift. So I used an alignment algorithm to minimize displacement between them before training.
In an other type of experiment, I recorded very large images. To save them, we downsampled them before training N2V.
Can you spot a problem when training N2V in either of these cases?
Structured noise
How can structured noise affect N2V performances?