147 - Getting started with Google Colaboratory for deep learning

Most of us cannot afford to purchase our own Nvidia graphics card for deep learning. Luckily, Google makes it possible for us to get started with deep learning using Tensorflow and GPU/TPU.
This video provides a quick overview of Google Colab and walks you through the process of mounting Google Drive and getting started with deep learning using GPU acceleration.

[@ShreyasKallingal]

To answer your question about importing python files, I think you can only copy paste the file into a cell if you want to directly edit it in the notebook. If you just want to run it via command line in Colab, simply run !python [path_to_py_file]. Hope this helps!

[@sabaal-jalal3710]

Thanks, Great explanation I learnt a lot from your channel :)

[@deylightmedia3266]

Sur can I run part of my application in colab and part on my computer/microscope/camera

[@surajshah4317]

Thank you sir Sir can you make more examples on multiclass segmentation

[@dibyakantaacharya4104]

this is amazing ....keep it and give explanation like this and new things ...thanx a lot for video

[@DigitalSreeni]

It's my pleasure

[@samarafroz9852]

Great job sir

[@epignosisal7107]

Thank you sir for the detail explanation. I observed that my data are not saved on colab, so I always upload them anytime I go offline or restart my system. How do I solve this please? thank you

[@DigitalSreeni]

I don't know what you mean. All my data on colab is always there, on Google Drive.

[@arungopinath1998]

If you need your data to be saved, please make sure data is saved to drive itself. Otherwise after each session data will be cleared, because google is providing temporary storage for the colab and its only active during the runtime. So consider saving results and other reusage stuff in the drive itself during the process. Hope it helps for some user I believe.

[@khaledf3977]

Very nice tutorial Thank you for great explanations

[@DigitalSreeni]

You are welcome!

[@talkhan5604]

Hey Srini! I am performing some analysis on H&E images and I am looking to construct cell geometry by first segmenting each cell. I'll get x,y coordinates from the 2D images but I will need 2D slice series (or Z slices) to get the z coordinates. How would one keep track of each cell across different z stacks? Also , do you have any idea where one might be able to get 2D series data (3D)?

[@DigitalSreeni]

In order to keep track of cells across different z stacks you need to write a tracking algorithm. This is not trivial. We are building some free modules (at my work) to do exactly this task. If interested you an sign up for www.apeer.com (it is a free cloud platform for image analysis).

[@talkhan5604]

@@DigitalSreeni Thanks for getting back. Apeer looks like an interesting platform. Can you specify the name of the module?

[@talkhan5604]

It would be great if I could get my hands on some 3D H&E images as well. Do you have any idea where I can find any datasets? Thanks