Dear Instructors,

This is most frustrating course in all of your courses so far. The instructions were completely misguiding the candidates from YOLO implementation onwards. All along you presented the course very well. But when come to most important topics, we had to focus on syntactical errors. But we are supposed to spend time on understanding the algorithms at this level. Dont know why this 180 degrees turn taken by you. If you intentionally designed this course then fine. Otherwise, you should seriously think about rework on the instructions. Few links to hints were taking to some pages in github with just folders.

I am sure , many learners here have such same opinion. I can see this in the forum postings.

From YOLO onwards, you were not giving the big picture of the task. This is confusing. We are lost, where we are heading by the mid of the assignment.

With all due respect to your highly precious time, I request you to enhance the assignment instructions.

All motivation I got from previous course, losing because of this course.

Personally, I feel YOLO easy to understand, but instructions were misguiding and confusing the candidates.

This is my honest feedback, as I very much like this course. I am going forward for the 5th course in this series.

Last but not least. Thank you for making this high quality knowledge made available for public with easy access via Coursera.

Amazing! Feels like AI is getting tamed in my hands. Course lectures , assignments are excellent. To those who are not well versed with python - numpy and tensorflow , it would be better to brush up.

This course is definitely tougher than the first three courses. Challenging but worth it.

Great course for kickoff into the world of CNN's. Gives a nice overview of existing architectures and certain applications of CNN's as well as giving some solid background in how they work internally.

Great lectures but the programming assignments feel as if it is testing your proficiency with tensorflow which is neither formally covered in the lecture nor the most intuitive framework to understand so you'll spend so much time digging through convoluted tensorflow documents and qna and whatnot to debug your codes that you would rather learn tensorflow formally first and then take this course and still end up finishing it faster than only going through this course only but it is only the programming assignments that basically assume that you are already familiar with the tensorflow framework so if you are only going to go over the video lectures it gives a great overview of how CNN works and many useful algorithms which can applied to a assortment of situations

This is very intensive and wonderful course on CNN. No other course in the MOOC world can be compared to this course's capability of simplifying complex concepts and visualizing them to get intuition.

Great Course Overall

One thing is that some videos are not edited properly so Andrew repeats the same thing, again and again, other than that great and simple explanation of such complicated tasks.

Theoretical material was great as always. However, programming assignments were poorly commented in some cases which results in unnecessary confusion.

I really enjoyed this course, it would be awesome to see al least one training example using GPU (maybe in Google Colab since not everyone owns one) so we could train the deepest networks from scratch

When I am doing the programming assignments, I felt that some part were quite difficult since I had no background in neither Keras nor Tensorflow. It was helpful that in one of the previous courses there was a tutorial for the basics of Tensorflow. But for Keras I felt that there is a gap between what I have and what is needed for the assignment. So I would suggest a more thorough tutorial for Keras. Maybe several short tutorials talking about the implementations and ideas of Tensorflow & Keras may help a lot.

Assignments are very bad explained

To set the context, I have a PhD in Computer Engineering from the University of Texas at Austin. I am a working professional (13+ years), but just getting into the field of ML and AI. Apologies for flashing this preamble for every course that I review on coursera.

This course is the 4th in a 5 part series offered by Dr. Andrew Ng on deep learning on coursera. I believe it is useful to take this course in order and it makes sense to study it as a part of the series, though technically that is not necessary.

This is one of the best courses to take if you want to understand the basics of Convolutional Neural Networks. CNN is a technically-difficult-to-understand, still-evolving field of Neural Networks, and it has thus far found remarkable uses in the field of computer vision. Dr. Ng really exposes us to this cutting edge research, by explaining research papers, starting from its 'inception' to work that was published just two years ago. There are several aspects of CNNs that are difficult to understand, including the very basic "convolution" itself. Much of that is made clear in his video lectures, which explore (and explain) a wide variety of Network Architectures in good detail.

The instructor videos are very good, usually 10 min long, and Dr. Ng tries hard to provide intution using analogies and real-life examples. The quizzes that accompany the lectures are quite challenging and help ensure that the student has understood the material well. As with the other courses, the programming exercises are the best part of the course. You get to practice, (1) decoding hand signs, (2) face verification, (3) face recognition, (4) single and multiple object detection (of cars in a street) ... All these problems are actual, real-life projects, which are extremely difficult to solve. They help the student practice the strategies and also provide a jump-start for the student to use the code for their own problems at work or in school.

Overall, this is an excellent course. Thank you Dr Ng and the teaching assistants, Thank you coursera.

Great course, easy to understand and very useful. The explanations are very clear, as is expected from the professor. The purpose of the course is for you to have a practical comprehension of CNNs, it will give you the necessary tools to implement you own networks, but it will not get into the specifics of each model. Nevertheless, all of the resources are referenced, which makes it very easy for you to dig deeper on any specific topic covered on the course.

Just horrible programming excercises, grader does not work, lost hours of life trying to modify random stuff only to find out that a) I need to copy paste my solution to an empty notebook, or b) that the "correct" value and the instructions are actually incorrect and could never get graded correctly. The content is very rough, the videos contained embarrassing outtakes for several lessons now. If you noticed Andre Ng repeat himself or the video go to black, those are not streaming glitches--those are actually in the content you are paying for. The transcripts are machine translated and contain all kinds of misheard words. If you are not fluent speaker and rely on the transcripts, you will not take away too much from the videos. All in all, the content is sort of interesting, but the delivery is horrible. Also, it is far too long, there is some random stuff that does not mesh with the rest in each week, just to make barely enough content.

Great course. The only improvement I'd wish is to get a better introduction to the concepts of Tensorflow and Keras.

Homeworks are too canned and do not promote deeper understanding.

This is the hardest course so far in the Specialization. I don't know whether it's me who is lacking the knowledge to fully comprehend the material, or it is because the material wasn't presented well enough. The experience was quite frustrating, and lost the motivation that I had been carrying through the previous 3 courses.

I appreciate Andrew's efforts in delivering the lectures with the aim to present most complex materials as intuitively as possible, but still I feel, because of the complexity involved with CNNs, the material need more explanations, and maybe extending the course duration to 6 weeks will provide more opportunity to the instructors to present the matter more clearly.

Quite remarkably, the programming exercises fell short on my expectation. It was not useful more than just a simple syntax-check exercise for TensorFlow/Keras/Python. It gives a false sense of achievement, when, in fact, I feel myself far from learning the theoretical underpinnings of the various algorithms explained throughout the course.

To conclude, by completing the course, I believe that you surely would get to know the vocabulary used in CNNs and would be aware of the algorithms developed in the domain, but you won't learn enough to implement these algorithms from scratch yourself.

Content is great, but videos could be trimmed to cut retakes. A big issue is that guidance for programming assignments abruptly drops off from extreme hand-holding to being thrown in the deep end.

Some parts of the course seemed incomplete to me, wanted more information on why things work exactly as described. Last week assignments have a number of uncertainties/bugs.

Fix the grader issues. Or at least give a bit more feedback from the grader. "You didn't pass, please try again" doesn't help. I'm sorry, but the last assignment of week 4 was really annoying. Even when you don't make mistakes.

Great course. Easy to understand and with very synthetized information on the most relevant topics, even though some videos repeat information due to wrong edition, everything is still understandable.

Wonderful course. Covers a wide array of immediately appealing subjects: from object detection to face recognition to neural style transfer, intuitively motivate relevant models like YOLO and ResNet.

Bugs in the programming assignments drive you to the point of giving up.

Out of the four courses I have taken in the deepai sequence this is the best one! This course got to the heart of the methods that researchers are implementing and also dropped you into the programming using Tensorflow. As noted by some other reviewers there are places where more instruction could be helpful, but I felt that this course obtained a good balance between information and challenging and also between concepts and hand-on implementation.

A couple of the prior courses could be merged and then this would be the 2nd or 3rd course in the sequence which would be much better in getting students to complete the projects and all courses.

Very good however most of the effort is applied in learning and applying programming (tf, Keras) than actually thinking about the DL models and practicing different scenarios.