糖心vlog官网观看

SUPER!!

Great

Good

I enjoyed the course and believe I learned a *little* of the material presented. One thing that I'd find helpful in the programming notebooks for the exercises is to add a little more descriptive material, either in text or code comments. I was lucky that I was able to complete the exercises, but often they required adding "print" statements to understand what was going on. I generally found the optional labs to be less valuable since they either couldn't be meaningfully executed, or presented contrived random results that were not very meaningful (see comments in https://deeplearningaigans.slack.com/team/U01BR86L13M for example).

In my opinion, all those `optional` papers just add unnecessary buzz to the studying process. If you think some particular paper is something really important, then better to do a video about this with explanation. Information should be presented in a structured way for better contribution to students intuition about the matter. Honestly, after the second course I feel a bit dizzy.

The instructor can speak a bit slower. Also, she can select common words instead of some informal or rarely used one. This course is not taken by people living in English speaking countries. The speed of the videos can be changed any value between [0,1]. For example, 0.75 speed for this instructor can sometimes become a bit slow.

Again I love the content, the information and everything in it. I dislike the "difficulty" of the exercises. Yes, the content in it is great but passing them does not necessarily mean you understood what you're doing.

I think this course has more advanced "tricks" and models that are supported with fewer assignments, which could be one shortcoming of the course.

Excellent understanding and practical experience, however the last assignment could have gone more ahead to semi final generated images

StyleGAN part is awesome although fairness in AI also took a lot of time which i didn;'t expected

Excellent course. Week 2 could have been better by talking more about Machine learning bias

Interesting subject, nice presentation, assignments are not intuitive

Good, but coding exercises could have been better

I think there is not enough practice

The relatively low rating comes from an average: week1 (5 *), week2 (1*), week3 (3*). The part of FID and other ways to evaluate GANs was useful and interesting; week2 on bias not so useful and felt almost out of place; week3 on StyleGAN interesting but a bit too high level. If possible, in this course the teacher managed to speak even faster than course 1 ... I'm missing the calm tempo of Professor Andrew Ng ...

Valuable but also far from perfect. In week 3 focused on StyleGAN, programming assignments show its structure but nothing further. I feel disappointed a bit as we didn't use StyleGAN to generate anything. I hope next course in specialization will further explore image creation and meet my expectations or it may be too difficult to code on my own and read all the included papers (as homework).

IN week three, most of the component of stylegan have a clear explanation, that is good. But it lacks the overall code architecture, how to link the generator with the discriminator in trainning process? how to stable the progress trainning in styleganit's important to get intuition about how stylegan work.

There are concepts which should be explained with more details, such as the content of StyleGAN (Week 3). The instructions of the 2nd week - notebook are not clear. Nevertheless, the course provides a good first approach to the state of the art of GANs.

the assignments where not that helpful, even tho the comments where a course on it's own, but when solving the assignment it may take you 4 hours just to learn the way the function works, which is the biggest issue in pytorch and scipy

You have to really love GANs, or have a real immediate need for them, to enjoy this course. All the earlier DEEPLEARNING.AI courses were pure joy, and not as much of a grind.

Week 2 is pretty raw - much reading and few explanation within lectures. After that programming tasks look like game "guess what to do to pass".

Lecturer speaks too fast.

The references are old, the course is in need of a fourth week with updates. It is good but I felt lack of recent information and real practice as happened on course 1.

too easy

The material is great; however, after the completion you lose the access to assignments and the only way to restore it is to subscribe again.

This course will have a minimum of mathematics explaining the work of GAN