#6 PyTorch Book Thread: Sunday, 3rd Oct 8AM PT

Note: This is a wiki, please edit it & add resources!

:tv: YouTube Link: PyTorch Book Reading - 6. Working CT Scan Data in PyTorch, Classifying Tumours. - YouTube

He y’all!

This is the new thread for our 6th session on PyTorch to discuss the chapter, host a Q&A, and anything else you would like to ask.

For #6 we’re at the exciting part where we can start diving into an end to end project. In this session, we’ll learn how to load CT Scan dataset using PyTorch and also learn how to build a classification model for the same.

:round_pushpin:Link to sign up
:books: Resources:

4 Likes

Interesting session covering the BTS of WandB :slight_smile:

Does LeCun really reply to the tweets us mortals make??!! :stuck_out_tongue:

Yes more homework would be awesome :smirk:

I think having the power to even reach out to these fathers of deep learning is a great power that twitter has given us. We should always ask, you never know :stuck_out_tongue:

3 Likes

Girijesh’s blog post on chapter 6 summary

2 Likes

Ask any Questions here please :slight_smile:

Talk’s slides available ?
also please share blogs link you discussed in start thank you

Here’s the slides:

Here’s my blog:

Hope this helps :slight_smile:

5 Likes
3 Likes

Data augmentation to increase the size and add more variations may be?

1 Like

We could normalize the data

We could crop into regions of interest if that is feasible

We could use affine augmentations

2 Likes

Something like boundary box?

1 Like

I’m not familiar with 3D images, but I would try:

  • normalize the images
  • apply some preprocessing steps
2 Likes
2 Likes

what does it mean @functools.lru_caches mean?thx

Is it a check to ensure the radius is not off by a lot?

Since there are 4 entries in CandidateDataTuple what is the field based on which the sorting is done in getCandidateInfoList function?

clip is a function in numpy I suppose which restricts the array elements to a particular range

https://numpy.org/doc/stable/reference/generated/numpy.clip.html

Can you explain again why function caching is required?

1 Like