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.
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
by Jamie K•
Lots of new concepts in this course. I liked the literature review sections and the fact that Andrew starts to show you when it makes sense to pull someone else's model down and use that rather than building something from scratch. The programming exercises were also pretty good - I had to think in a number of places though they are still a little too structured for my liking.
by Najeeb K•
A great course providing in-depth theoretical understanding of Convolutional Neural Networks and state of the art model architectures for various Computer Vision tasks. I have been doing Machine Learning from past one and a half years but the course content still gave me wealth of knowledge in a structured format that I yearned for so long. Thanks Prof Andrew and the team! :)
by Antony A•
best course in world or unvierse to understand the basics and complex details of convolutional neural network .i would give an oscar for this course . I was so woried about the complex diagrams that i saw in internet about CNN but this course made it look very easy i was totally suprised how complex details were explained in simple manner .I would recommend this to everone .
by Manjit P•
This course covers lot more material and it is more application oriented compared to last three courses. I had to spend lot more time and effort for this one. Also, there are some bugs during submission of the assignments. There is enough discussion about those but I hope Coursera takes care of those in the near future. Nevertheless, I always enjoy Prof. Ng's lucid lectures.
by Utkarsh M•
This course was something different. Earlier when I started The Deep Learning Specialization, I was not interested in any particular application of Deep Learning, but this course gave developed interest in CNNs, so much so I'm seriously considering and planning to pursue my master's in Data Science. I wanna thank Andrew Ng for such great lectures, he has truly inspired me.
by Yash M B•
This course has given me everything that one can expect to learn from the field of Image processing models like CNNs, Deep Convolutional Models like Inception, VGG-16, VGG-19, ResNets, etc. Other topics were also learned that included me applying these concepts into real-world applications like the neural style transfer as well as the object detection and face recognition.
The teaching style of Dr Ng is excellent as usual. He is able to take a complex topic and make it easy to understand. I found this course more challenging than the others in this specialization. It does require a bit of tenacity in order to finish the assignments. This is usual when coding. So don't give up and be sure to search the discussion forum when you hit a barrier.
I am really appreciating this specialization. The only thing that I would change is maybe focusing less on the matricial operations required e.g. in the loss function computation, and more on how to use Keras/TF at a higher level; at the moment, it would still take me a lot of time figuring out how to build a nn from scratch, or use an existing one, with these frameworks.
by Erman N•
This course is amazing. I strongly recommend everyone willing to build a career in machine learning to start here. I was really skeptical at the beginning. As a Ph.D. student in the computer vision field, I was looking for a course that can simply explain the science behind most AI courses. Now, I can say Andrew nail it, the course was far beyond my expectations. Thanks
by Esteban C•
Very good in-depth coverage of conv NN.
Just one little thing, week 4 Notebook assignments:
In style transfer code is not well explained how the train is actually working. In this case the input is set as a Variable instead of a Placeholder and this aspect is not mentioned or explained
In face recognition I still don't know how triple loss function is used during training
by WALEED E•
This course was the best I have ever taken. It gave me a big boost to carry my PhD research in robot vision with confidence of understanding what is happening all over the network and comprehension of one of the pioneer papers published in discussed in classes. Coding directly after finishing each week was the best to go to practice and apply all this knowledge gained.
by Ayush K•
Quite lucid and good introduction to CNN for beginners to intermediate level. I specially liked the links and discussions about different papers along the course that Andrew recommends to read. For some who has just hear about CNN, but knows about basic NN, this is a really good course to learn main things super fast and then proceed into their own personal topics.
by Kseniia P•
Amazing course with clear explanations of how CNN works. Andrew gives you intuition and understanding of convolutions, pulling, padding, and explains the foundations in great detail, so you can understand state-of-art approaches and are ready to get hands on it. Thanks to the assignments' structure, you don't ever have to waste time on debugging irrelevant issues.
by Teye B•
I love this course. I only wish there was an opportunity to go step by step from looking at images, creating the dataset from the images, creating labels, applying a model, and then testing. This would help to answer a few questions that I have. However, when I read the papers recommended, I assume many of those questions will be answered, such as : why max pool?
by Umendra C•
Best course on deep learning for computer vision! Convolutional networks can be tricky to understand, but Andrew has presented the material in a very easy to understand format. He starts with simple ideas and concepts and then build on them in an intuitive manner. Highly recommended course for anyone who wants to understand the deep convolutional neural networks.
by Michal M•
Excellent course. Time well spent.
Simple explanations of difficult concepts.
I was able to download yolo v2 in pytorch, reconfigure it to use CPU on my Mac, and get it running on my webcam in 1h after completing Week3 assignment.
Told all my friends how awesome the course is.
Keep up the fantastic work.
Super stoked for part 5!!! and learning GANs and RI afterwards.
by Akshay M P•
The best course on Convolutional neural network I ever had! This course packs in a lot of information delivered in a very effective way. A glimpse into the development of various CNNs gradually builds up into state-of the-art implementations of very deep CNNs. The coding exercises gives the right amount of exposure to the frameworks and tools used in the field.
by Peter D•
Great course from Andrew Ng, as always. The videos are superb in explaining some of the more recent algorithms and trends. And they provide good intuition on how to use them in your own work.
The only (minor) remark is that the exercises might not be that challenging for those that already have done some ML programming in the past.
But overall still 5 stars!!!
I was always curious about the "CNN" concept every time it emerged in the news. Thanks to Prof. Andrew's mild explanation, now I get a straight intuition into it!
The assignments were very amusing in this section. It was not hard to get a pass with the help of forums, but understanding every step is more important I think. So I will come back to practice more.
by MONIL J•
This is the best course for beginners as well as intermediates, to learn from basics and scratch up to the advanced of CNN. In this course, the fundamentals as well as all different CNN architecture and Face validation, recognition and neural style transfer has been covered and explained in very easy language.
Thank you Andrew Ng for such an amazing Course!!!
by Sherif M•
Again a great course by Andrew Ng and his great team. Convolutional neural networks are the reason for the recent Deep Learning revolution or let's say better renaissance. Andrew does a great job in explaining the theory, math and application fields of CNNs while also telling about the history of recent advances in CNN algorithms and architectures.
by Jaime M•
As in previous courses, Andrew made understandable complex and abstract content. This course is by far more challenging than the 3 previous ones. Maybe not at the assignments as we make use of facilitating frameworks and helper functions, but to really follow what is happening behind... its another level compared to previous courses on the specialization.
by Ammar A•
It's thoro and concise... the best part is the assignments are interesting and we learn quite a few things in the course which talk to newbie perception of DL... so things like Face Verification yep now is the right time not only to learn how to implement but also learn the quirks & features of CNN's... Course4 all those efforts are indeed paying off...
by Adrien S•
Great overall course, keep teaching please ! I learnt a lot. I have a Ms degree in Machine Learning but we didnt had the time to really learn about Deep Learning. I feel it was a great introduction to the field and I feel confortable now to get more in details about everything and read papers etc.
So thanks for that, and I can't wait for part 5 about RNN
by P M K•
This was a really good course to see mini projects getting executed. It gave quite a lot of practical insights working on the problems. The only issue was that week 4 assignments had some bugs in code comments due to which people spend quite a lot of time debugging causing unwanted waste of tine and frustration. Please correct the errors.