Great course, but assignments are from trivial to insane.

I would like to learn the ideas, but spending like 10 hours "learning" names and grammar of some packages, that will probably be outdated in next few years, is nothing fun for me.

What I really liked about the course was the actuality of the paper. However, I would have thought it absolutely necessary to explain the BackProp for CNNs. Also the grader problems in the last assignment force me to subtract two stars.

Face recognition notebook has a bug, I passed the grader but the function triplet_loss returned the wrong value in the notebook. Several other people have had this problem despite the fact that the notebook was supposed to be updated.

Assignments do suck.

If model cannot be run locally there is no way to debug it. More test cases that would cover most common mistakes would be quite useful. Otherwise the only way through is to burry into forum topics for hours.

The lectures were messier compared with the previous courses. Lot's of problems with the grading tools. The content of the course is great, so I would recommend it to others, modulo warning the others about being more patient :)

This is one of the weaker courses in the specialization. I wish it had gone more in-depth. It's so far the most complex problem and I don't feel like it has gotten the same attention as the basics did, in the other courses.

Covers interesting material and practical problems, and tries to get the student to implement useful tools, but there is a large disconnect between the understandable theory and frameworks used to implement the solutions.

Good course, but with the conjunction of the poor quality of the Coursera interface, video quality, the price does not feel like a great bargain. Still I feel confident I can be efficient after following this course.

Exercises were purposly difficult because of obscure API documentation and quirks (not because the problem itself was difficult). Good school in debugging, I personally disagreed with it (V3 if I remember correctly).

Enjoyed the course but the programming assignments weren't well designed I think. They were more about debugging than applying what was learned. I preferred the assignments of the earlier courses of this curricilum

The Face Recognition lab just don't make any sense, the expected outputs are the ones in the Face Recognition for the Happy House. And it made the exercise very annoying! Despite it, the course is really good.

The production of the course felt rushed, there are numerous clipping issues in the videos and a major bug in one of the assignments. Also, for such a key topic to be covered in only 4 weeks felt very shallow.

Good explanations of the material but bugs in homework assignments and better explanations of tf usages is required for certain assignments. A refresher of tf via an additional assignment would've been nice.

Good insights on the YOLO algorithm as well as in Siamese networks and triplet loss. Miss some more deeper understanding both in the lectures and the assignments, but I totally recommend the course anyway.

very good topics discussed ,facial recognition and facial verification assignments do not do justice to the complexity involved.practical knowledge gained is less compared to other modules prior to this.

The knowledge is good, and the techniques taught are valuable; however, having to use a deprecated version of TensorFlow is annoying and a lot of this will have to be re-learned to be put into practice.

Course is very informative.

Unfortunately, unlike other courses in the spec, there were quite a few bugs in the notebooks and they took quite a while to load due to the sheer weight of the models loaded.

Face recognition is a bit oversimplified, there is more to it that a simple accuracy metric. Priors are involved, which are included in the NN training, but should really be disassociated in evaluation.

In contrast to course 1 and 2, I've found the videos to be rather shallow (no pun intended), in the sense that, in my opinion, they haven't explained thoroughly the techniques' underlying mathematics.

The information in the lectures was brilliant. However, the coding assignments don't really test your understanding of the course, rather your ability to piece together the authors previous code.

This is a very interesting and well structured but the assignments in week 4 got alot of bugs, grading gives zero points for the right ouput (according to the notebook) and ten for a wrong one...

The course starts normal but suddenly gets very confusing from the start of week 2. Also it gets a bit difficult to understand things later on. Otherwise its very good course and i enjoyed it

The Labs are so confusing and I know the theory but the labs aren't good enough. Instead of having lengthy Labs we should be doing labs after 3 or 4 videos that way things can stick better.

Overall the specialization this course is the more complex, not only regarding the main concepts I think that the assignments are hard and will be usefull have more context about tensorflow

So far I was very enthusiastic about the courses but this one is rather disappointing. Unfortunately, the video editing is very poor, if done at all, which make listening somewhat annoying.