The grader in the final assignment was giving some problems even if the answer was as close as last decimal digit.Else the course was excellent ,it gave excellent insights to different architectures used in CNNS and its working.

Course content was really upto the mark with the current trends in convolutional neural networks, its the best choice for those with pre-gained knowledge about machine learning and have and idea of updating their skills in cnn.

As usual, the content is top quality. I did however notice several times in the recordings where Andrew tried to state something one way and then went back and restated the same thing - it seems like this should be edited out...

Really nice course of convnets. I think the intuition behind them could be more explained. In my opinion, the artistic part of the this nets should not be in this course as there is plenty of more important knowledge to adquire.

Thanks. Great course :)

There're some mistakes in assigments. Check the course forum when you're in trouble. The course forum is very helpful. Thank all the students and TAs who post their questions and solutions in course forum.

A very good course, but probably the weakest of those so far in the Deep Learning specialisation. Got better towards the end, but the first couple of weeks felt like some topics were not fully explained in terms of relevance.

Good course but I would really recommend doing a deeper explanation on backprop of CNN's. Also I think the explanation of YOLO is not arranged well. ultimately it gets there but when it is first introduced a LOT is left out.

Great material, but there was a bug in the grading of the final problem. To work around the bug in the grader, everyone is modifying their program so it doesn't match the expected results but nonetheless will pass the grader.

I think that you should give more details and instruction how to implement our own images, how to perform Transfer learning to the ConvNet for improving, how to train the network myself not only load the pre-train network...

Very Interesting

As usual, it is very very good from theory point of view. Practical examples are also really interesting.

Do not expect to be autonomous after the course, as you won't be able to use Tensorflow or Keras alone.

This course provides great introduction to CNN. It is an eye opener if you didn't know about CNN and it will take you to a level where you will be quite comfortable with CNN formulas and know a bit about how to implement it.

I learned a lot about CNN. Good programming assignments. Unfortunately the quality of this course was not as good as the three courses before (Grader Bug in last programming assignment, see discussion forums for workaround).

Вообще четко, еще бы отредактировать субтитры на английском. И цены не будет. Звезду снимаю только за то, что когда описывается модель с распознаванием и позиционированием начинается сумбур и надо по несколько раз смотреть.

Great lectures and exercises but the leave-behind study materials (lecture slides and notebooks) could be a little more helpful, and the frequent server disconnects have forced me to do most of the programming work offline.

A lot of useful information.

However, video editing is awkward sometimes plus Andrew's voice sometimes gets down to wisper which is hard to catch.

So good course but there is defenitely some spave to improve presentation.

Loved the content. Covers the fundamentals of ConvNets wonderfully. Need a lot more clarity on the use of TensorFlow and Keras in building these systems. Particularly the programming exercises felt a little obscure to me.

Although the course is comprehensive, at times, some of the concepts were skimmed over, especially in weeks 3 and 4. Topics like UNet architecture and Neural Style transfer felt rushed.

Overall, the course was excellent.

Great Course! I did the sequence model course before CNN. CNN programming exercises are challenging than the other four courses but could solve it with the support of discussion forums and hints provided in the exercise.

the general concept is important to understand the data flow from one block to anther block. this will give me big picture to learn convolution Neural Net works , thanks all content and support on this course.

-jianguang

I had to watch the videos more than once to grasp all of the material. The exams were harder, and no partial credit was given nor specific indication of missed questions on the quizzes. Great material, very applicable.

The lectures are awesome as usual, but I would prefer less guided exercises with less fancy content but more I don't feel like I have implemented anything myself unfortunately. Still, big thanks for the great pedagogy!

Not as great as the previous three.

Prof Ng's explanations are flawless and awesome as always but the programming assignments had more issues but I think by the time others take this course all issues will be resolved.

The judge system of last week's work has some problem, but the course of Andrew is still worthy to learn. Learned a lot about CNN from the course, but still need to read more papers in order to step into the field.

This is a great course with a lot of useful content. The only reason I am not giving it 5 stars is that it was too packed. I would suggest Andrew and team to convert this course to a 6 week course instead of 4 weeks.

Overall the Course is good but the only examples given in assignments or lectures pertain to Computer Vision and Image recognition/manipulation. Surely a few examples or discussion of other uses of CNNs is in order.