Great lectures and exercises in "digestible portions". The course explained the GAN basics first and then built upon that base knowledge in a gentle and well though way. You always think that by just reading papers and reviewing reference implementations you can master some new ML-area, but this kind of course is so much more fun with materials, community and support.

1. good introductory course for absolute beginners.

2. The instructor speaks very monotonically and it seems she is reading a script. Explainations could be a lot better.

3. A suggestion: Incorporate optional part in the lecture series as they contain more important and detailed mathematical explanations, which tbh i expected the instructor to cover in the course

This course is great view into GANs. The lectures often briefly review the basics of topics like neural nets or convolutions, yet still offer advanced (optional) lessons and journal articles to read.

I rated 4 stars instead of 5 because I could not find printable/PDF notes for the course, unlike some other courses.

Sharon does a great job of teaching concepts, and the course follows well from the Deep Learning Specialization. You will find that while the code exercises start out facile, you will require some help on the Slack channel, almost no code support is given in course (in contrast to Nanodegree programs).

Really excellent course I didn't have much problem following along with the lectures although some coding exercises could be made clear it was a great learning experience. There should be some more notes visualizing the generator architecture as the number of layers in the generator are not clear.

I think this is a bit too basic, there are some areas where i believe some more maths and theory might be appropriate. IE specifically the video section prior to W4B programming exercise with the latent z space manipulation to target disentanglement of features.

Very good course to understand key concepts of GANs. However, I think it would benefit from building small blocks of GANs at a time and see how we end up with a functional model, instead of giving us a notebook where we have to fill few lines of code.

Good for basic GAN knowledge. Good for Pytorch knowhow, if you are new to it. Concepts are explained in easy to understand way.

More mathematical explanations on probability distributions of real and fake images, Their distances would have been better

FYI to course creators...

Almost without exception, the correct answer to the quiz questions was the longest answer. I know the quizzes are not graded, nevertheless the consistency of this pattern reduces the value of the quizzes as a learning tool.

Perfect course. I wish we would have more detailed explanations or visualizations of the steps of GAN architectures. Also, W4 conditional GAN assignment grader could't run my code on the first even it has passed the test in the assignment.

I don't like that it skips a lot of mathematics behind the concepts. The programming assingments were nice. I would really really like it if they also added mathematical explanation behind the concepts taught. Otherwise it was a nice course.

Good course!

My suggestion is that we need to add more detailed examples (with numbers) (not just shown in the assignment) immediately after the introduction of a model, which will significantly help the students to understand the model.

The course was very good, only complaint is that assignment w4b was a little vague, in terms of comments on the code and even the fact that no paper or explanation was offered in the course for in depth implementation of the algorithm.

Great material and instructors. Enjoyed watching videos and taking assignments.

Assignments could have been more difficult if we assume most people have taken the deep learning specialization or are familiar with deep learning.

Thanks

I love the explanation and what you actually do in this course. However, if I were to use this to evaluate whether a candidate for a job can work with GANs in practice, I think the complexity for passing the exercises is too low.

Nice material, but the assignments are extremely rudimentary (paint by numbers/fill in the blanks). Perhaps you could provide more advanced (even ungraded, if that's the challenge) assignments for folks that want them?

Thanks

The voice over was pretty fast and hard to understand, so had to do lots of sliding window in video to understand the topics. Otherwise this course is fantastic gateway to understand GAN and it's applicability.

The Teacher is awesome the way she explains the concepts through great examples. I wish the exercises were a little bit more handson and independent (most of the code structure is already there).

Great material. At times, I think there wasn't enough explanation to get the right answers for the assignments, I needed to guess at times and not completely understand what was going on.

Great intro course, the programming assignments were pretty weak in difficulty level, could have had less hand holding there. Excited to get into more high resolution GANs soon!

A little lacking in technical knowledge. You just get to build a GAN and understand bits and pieces about why it works in very simple terms, little mathematics involved.

Lectures were clear, focused & informative. Assignments were well-formed with extremely helpful hints & explanations of the code to be completed. Nice work!

Great course to start building GANs.

I wish more math was included. I realize the math behind this is very complex, and not everyone wants to know about that.

Great examples. Wish there were more reading material that bridged the gap between the papers (very detailed) and the slides (good for exposure to material)

An excellent course. The only area of improvement I can think of would be to get better intuition on the tensor shapes through the model building code.