I haven't seen everything, but here's a review of the course first..
[Advantages]
1. It organizes the prerequisite knowledge well at the beginning
2. The lecture name is CNN, but it is not limited to CNN, and it explains the basics of deep learning (SGD, Backprop, etc.) in detail, so it is easy to understand even if there are some difficult applications later
3. The image preprocessing is also detailed, so even those without basic vision skills can try it (I also recommend listening to Professor Kwon Cheol-min's vision lecture)
4. It is not just a simple CNN image classification, but also a detailed explanation of how CNN has developed recently
5. There are many pictures in the lecture materials for easy understanding
[Disappointing points]
1. It is a TF-based lecture, but torch is also...ㅎㅎ
[Overall]
5 points. Those who are new to deep learning in the image field should definitely listen to it, and those who are just new to deep learning should also listen to it because the basics of deep learning are explained in detail. CNN itself is honestly being used not only for images these days, but also for NLP and predictive modeling, so it is good to understand CNN deeply and utilize it.
Hello, I am currently studying artificial intelligence as a major. I will not lie, but the lectures are conducted according to the university level curriculum.... The difficulty level is about 90% of the university level(?). It is not as deep as a master's degree, but this lecture definitely covers everything from beginner to intermediate level. And apart from the difficulty level, school classes end with just reading PPTs, but this lecture was so good because I could follow the code line by line. Sometimes, in some areas, he explained at a level that made me think, 'Huh?! They go into this much depth!?' I was surprised. (I was surprised because the points and explanations that the professor made were almost the same.) The regrettable thing is that the parts that are the pinnacle of modern machine learning, such as transformers and attention, are not covered yet. If this lecture comes out, I think Professor Kwon Chul-min will almost become the king of domestic machine learning private education. It is the most satisfying lecture among the lectures that I paid for at Inflearn.
I thought I had a general understanding of deep learning while working on the project, but this lecture made me realize that I had been doing it by the rule of thumb. I keep listening to it whenever I have free time during the day. It is very helpful, and if an advanced course comes out later, I would like to take that lecture as well.
I've watched about half of it now.. When will the advanced lecture come out? I feel dizzy, so please make it before I listen to it 100%. I want to listen to it quickly.
CNN is a relatively new field if you have done a little bit of deep learning, so you may think you know it, but it is harder to respond flexibly than you think because you don't know the principles and use it.
Whenever the company needs it, I search for github or quickly combine the functions I need at the time without knowing the principles.
While doing it while doing other work, I kept putting off studying tensorflow2.4, keras, and kaggle, thinking that I would study them someday. However, I remember successfully applying object detection to robot movement in a project I took the last lecture, and being recognized as an expert(?) at the company, so I am taking the class to organize CNN again and look into it in depth.
The class time is not short, but it is not bad because the source code is explained in detail, so I think it would be good to quickly look through it and mark it separately and listen to the parts I need in more detail later.
---------------------------------------------------------
I am adding this to the review I left before. I did other work at the company, and then the computer vision department was created again, so I listened to it from the beginning again to review it.
When I listened to it again, I realized that I had a wider range of understanding of the parts that I had rushed through before.
This is not a lecture that you can listen to once and finish.
I highly recommend it again.