Lecture 6 - Image Generation using Generative Adversarial Networks (GANs)

Livestream Link: https://www.youtube.com/watch?v=ABaZ_tecZ3U

Lecture Timing: June 27 (Sat), 8.30 AM PST/9:00 PM IST

Notebooks & References:

What to do after the lecture?

Asking/Answering Questions :
Reply on this thread to ask questions during and after the lecture. Before asking, scroll through the thread and check if your question (or a similar one) is already present. If yes, just like it. During the lecture, we’ll answer 8-10 questions with the most likes. The rest will be answered on the forum. If you see a question you know the answer to, please post your answer as a reply to that question. Let’s help each other learn!

Office hours day 1: https://www.youtube.com/watch?v=16fwbhAEFFw


WOW, finally come to final week, learn the GAN, awesome.


Is the course going to roll out again for another certification session?

1 Like

Hey, @aakashns do every project need to be unique because i think someone has already worked on the dataset i was working on.

1 Like

@aakashns: Do we have to write the project summary in the form of a blog post or will a well written notebook also be fine?

This is the best thing, I invested my time in this Lock-down period.


So far went well.
I think I will be joining the next one also.

1 Like

I think thats not possible.
because most of the projects are uploaded at the last hour and most of them will be repeated

1 Like

@aakashns how exactly can we submit that we participated in the competition and how well we did?

1 Like

@aakashns Could I get some suggestions for classifying rectangular images in a CNN. I’m working with 4k images and I believe that they are too large. I was thinking on cropping them to the area to a 128x128, but I don’t have the medical knowledge.

While training a model on kaggle kernel the notebooks turns off after 1 hr and all the progress is lost any way we can prevent that??? @aakashns


just type something like print and stuff no need to run that cell and your session will not expire

1 Like

Faced the same Issue, but couldn’t figure out any way.

How the choice of mean and std 0.5 is done? Also is there any advantages over typical channel wise normalisation, in which the mean and std is taken from the distribution of training data.

@aakashns How well do CNN do with classifying more that 10 labels? The dataset that I’m working with has 1000 images for 39 labels.

@aakashns Could you walk us through how do you select the “stats” for normalization? Is it like a hyperparameter or do you run a algorithm to get the parameters?

1 Like

Are there resources anywhere for how to access old models we have saved to our profile?

how to covert random set of images to pytorch tensor.

1 Like

Are you talking about state_dict?