Specially in week 4, big gap between information taught in the lecture, and the last assignment.

poorly designed exercises

poorly designed material that all could be said in just 1 or 2 weeks

Some notable omissions of key information in instruction, but mostly the exercises suck

The content are good, but the audio is low and the exercises are not well documented.

very basic.

There's hardly any video time. We can finish the whole course in a day

HWs are not well and thoroughly-thought designed (Grader runs out of memory ).

Average content. The last assignment for week 4 was structured quite poorly.

coding assignment should be included otherwise it is easy to get certificate

expecting more advance topics instead of just using Keras.

The jupiter notebook exercises are a mess, pls, fix them

The last assignment is poorly designed and a real pain.

The practices were poorly designed.

Assignments are not well designed

Very little content, ridiculously repetitive. Missing info in the last assignment made me waste a lot of time. I wouldn't mind spending time on that assignment if it had been actually challenging. Course designers should consider putting serious time and effort into the preparation of the hands on materials. There is some useful information in this course, but I passed without learning much because of poor design. I still have no idea how to design a good model based on the data set, how many filters to use in each convolution, how many layers in total, how many nodes in the dense layer, etc. The instructor keeps recommending "trial and error", which is fair but I would expect the discussions to be a bit more thoughtful and deeper than that, especially given the price of this "professional certification". This course is falsely advertised as intermediate level, it is actually introductory level, albeit being too simplistic compared to some other great substantial introductory level courses on Coursera.

Honestly this course was a bit of a disappointment, didn't really learn anything that can improve how effective i can produce a neural network, on the assignments the hardest parts was coding tasks that was never explained in the course, like reading and copying files, and that has not really anything to do on how to build a neural network, some of the examples provided did not really worked (as the accuracy was poor and always over-fitting), for the last assignment i cheated as using image generator lead to a really poor results so i just skipped and train the model using directly the images.

My suggestion spend more time on what strategy are used to improve the model when the results are poor as the model under-fit or over-fit, otherwise it looks like you add and remove layers until you got lucky.

I am sorry for the 1 star but is time that deeplearningAI take course content difficulty a bit more precise. I know and I appreciate a lot what you are doing on spreading the knowledge of AI through many varied content, nonetheless if you put that the course is intermediate you should aim to make it intermediate. This is pre-introductory level course and I would have appreciated knowing that so I just skip it an go through a more advanced option. I still finished it as I already paid for it. It is still a well organized and fun course as usual, the problem is that this has absolutely no value in terms of "real world" examples as you put it...

What a disaster! I spent more time trying to game the scoring service into accepting a correct answer (and reading in the forum how other frustrated users managed to do it) than actually learning about tensor flow. Such a shame because there were some interesting topics. There is no TA answering questions in the forum either. It doesn't feel like any effort has been made to keep the course current and updated, addressing issues, etc. And the lab work was actually very repetitive - some were almost identical to previous ones.

Final lab was a total shit-show. Made you jump through hoops to run off and learn minor things which weren't covered in the course but were highly germaine to passing. Tons of technical problems with getting final lab to submit - reported by many users for over a year - never resolved by Coursera. But for the course itself, it was highly redundant - going into longwinded explanations over several videos of very simple things, often covered in the prior course.

I am very disappointed with this course. The lectures consist of 1-2 minute videos and 30 second lectures. The quizzes seem unserious and the exercises are data organization tasks and not TensorFlow programming, since the TensorFlow related parts can be solved by copying and pasting code snippets they give you in the Labs. Even the exercises in weeks 1 and 2 are the same, just changing a function call! I didn't expect something like this from DeepLearningAI.

Course is not properly structured. Transfer learning was using a very special case not the general case.

It is clear that course is just a collection for some scattered old videos and materials. You will realize that at the end of the course when you find that final assignment is NOT relevant at all to the material. Final assignment is a nightmare where there are no guidelines at all. There is no support at all from Coursera. it deserves 0 out of 5 not 1.

1) material is boiled down to no-brain

2) questionnaires have incorrect terminology (like method vs parameters)

3) with almost no mandatory assignment the value of certificate is dubious (see point #1)

4) Please stop marketing this course as about TensorFlow. it is all about Google's implementation of Keras and Keras only.

5) The code in course DOES NOT WORK with TF 2.0-rc. Thus student with 2.0 can't submit assignments.

The course overall was great, but several notebooks were really frustrating, *especially* the final notebook in the course! There are literally 0 text cells that explain what is happening, 0 information on the goal, 0 information on the data we are working with, is it dogs or human or hand pictures? Are there 2, 3, 100 classes? Literally 0 information. This notebook was super frustrating to complete.

On the first course I gave more detailed feedback why I disliked the course, and unfortunately those ideas and feeling are even stronger with this one. Summary: The course is way too shallow and puts focus on many different things rather than what it should have done which is build on the deep learning specialization and do in depth focus on tensorflow.

This course severely lacks depth, and the explanations are almost non-existing. Some theoretical background is completely missing and I wonder what the target audience is? Beginners will not be able to follow as many things are left unspoken and intermediate to expert learners will be left with a feeling of incompleteness. Big disappointment.

The discussion boards are unmoderated and none of the instructors are active. You need to go outside the scope of the program to get your code to work. Also their grading software often crashes because of THEIR code. I do not think they tested the class prior to uploading the materials.