2
votes
Is it better to train neural network with ideal data before resuming with non-ideal data?
If you have the full dataset (comprising both "ideal"/cleaned and "noisy"/real examples), then no, it is not necessarily better to pretrain it with clean data just to initialise ...
- 121
1
vote
Accepted
What is the channel dimension other than color representation in Conv2D? Shall I use Conv3D instead?
You can do whatever you want, but it won't necessarily achieve a good result. CNNs have the inductive bias that features are spatially correlated. If you simply stack different views and immediately ...
- 85
Only top scored, non community-wiki answers of a minimum length are eligible