I want to know how to setup CNN using Jupyter from the scratch. Also I would like to know how to add custom image datasets from local folder to CNN done in the above setup.
What do you mean with "from scratch"? Do you want to implement all methods yourself (that is, back propagation, convolution etc)? Or do you want to just use and train a CNN for your data? In the latter case, I would suggest to just go to the Tensorflow or PyTorch websites and learn from the tutorials. You can easily load your image datasets using - for instance - a PyTorch data loader.
Ferdin Joe John Joseph I would suggest to use PyTorch, you can easily use it inside a Jupyter notebook. You can find install instructions on pytorch.org, and also some tutorials on how to build CNNs.
To use your own data, have a look at this tutorial: