The course was very basic but interesting. However, there were some issues when submitting the assignments. And the virtual lab uses tensorflow 1.x instead of 2.x

It's very introductory and the knowledge may not stick. I think it is more beneficial to take a full deep learning course with TF as an add-on to the course.

Great content, excellent explanations.

But I couldn't run the notebooks without running into kernel issues, the programming assignments were a real hassle.

Decent enough but much too abbreviated and lacking the depth I expected from a deeplearning.ai course after taking their deep learning specialization.

Could have been a more elaborated course. This course mostly talks about how Keras functionality has been adopted by high-level APIs in Tensorflow.

This course gives a high-level overview to tensorflow keras api which is good to begin with but working on complex use cases would be preferred.

This course was not focused on learning tensorflow as I had hoped. Instead, it felt like an into to neural networks course using tensorflow.

I missed graded exercises. WIth only the simple multiple-choice questions tests it becomes a try and retry game to score as high as possible,

Too easy and lacks theoretical explanation, even though there are references and it seems that it lacks the explanations on purpose...

The quality of assignments should be better. There should be less emphasis on overfitting the data to 99% or 100%

Compared to the Deep Learning Specialization, this course falls short in course depth and coverage of material.

I know that it is an introduction but I would like more staff .

I am quite disappointed with this course. First, it should not been called "Introduction to TensorFlow" but "Introduction to Keras", which is a TensorFlows' (TF) API that entails a higher layer of abstraction. Basic data structures, estimators, graphs, etc. are not explained through the course. Second, video lessons are too superficial and lack of content. They remind me to those of the Machine Learning Crash Course from Google. That is, as an opener/introduction for Deep Learning (DL) are fine but they are far from being an essential training tool in DL (unlike the Deep Learning Specialization here in Coursera). Finally, content is too basic. This course requires an intermediate level, so students are supposed to be already familiar with basic DL concepts. I understand that this first course within the specialization is an introduction, but I just begun the next course (Convolutional Neural Networks in TF) and it is more of the same. Laurence is still working on the binary classification problem and only at the end he treats the multi-class problem. Instead, I was expecting to implement CNN models like ResNets, Inception networks, and applications like object detection or face recognition in TF (not in Keras). For me, it is not worth spending time and money for what you learn in this course. The good part is that, because videos are short and exercise are easy, you can finish the whole course in just one week (or less if you are 100% working on it).

Maybe I had unrealistic expectations following the original 5 courses from deeplearning.ai. I understand the target audience and need for simplification, BUT there are multiple outright wrong statements, that are unacceptable (will list below):

1 `Softmax takes a set of values, and effectively picks the biggest one, so, for example, if the output of the last layer looks like [0.1, 0.1, 0.05, 0.1, 9.5, 0.1, 0.05, 0.05, 0.05], it saves you from fishing through it looking for the biggest value, and turns it into [0,0,0,0,1,0,0,0,0] -- The goal is to save a lot of coding!` - no it doesn't do that, it takes n numbers and gives n numbers which sum to one and respect all original inequalities. and no it doesn't save time, you still need an argmax.

2 in the first course there's a linear regression trying to learn f(x)=2x-1. The course says you can't get it exactly because you don't have enough data. Of course you have enough data, 2 points are enough to describe a line, and that regression has a closed form solution. SGD with fixed LR is the only problem.

3. Immediately after, also first lesson, it says that sometimes loss goes up and that's called overfitting.

Those were just a few...really I understand it doing baby steps for developers without maths background, but I'm not sure this is doing them any favors..I've also showed these to a bunch of my colleagues and we were on the same page about it

No deep details for functions used

This specialization is false advertising. It does NOT prepare you for the Tensorflow certification exam. It’s especially disappointing after taking the fantastic specialization by Andrew Ng, and makes this specialization feel like a cheap cash grab by Coursera and DeepLearning.ai. This series of courses fails to prepare you for three reasons:

1 – The certification exam is done on whatever is the current version of Tensorflow (v2.6 as of writing). You can’t expect a specialization like this to update every minor release, but much of the coding is still on the v1.X version!

2 – The certification exam requires you to work in the PyCharm IDE. The IDE doesn’t even get a mention in this specialization and it is all done through Google Colab.

3 – The material is covered at a very superficial level. I was hoping to walk out of it feeling confident in using Tensorflow on novel problems, but I’ve barely learned anything about Tensorflow that I didn’t already get from Andrew Ng’s specialization. There’re a few minutes of lectures (some weeks less than 10 minutes). The programming assignments are either pathetically easy, or lack any guidance on what to do (seriously sometimes there’s no instructions at all, you have to guess what to do by the variable names), or both.

Save your time and money and go elsewhere to learn Tensorflow.

Code and exercises look like they were made in a hurry, with a lot of errors that have not been addressed yet, even after been reported about 3 months ago. No challenging practical exercise (just need to copy the code from the previous notebook that the instructor supplied) (maybe making the function print "Reached X% accuracy so cancelling training!" was necessary to fool the grader). Weak theoretical test. I had high expectations, and now I am disappointed with this deeplearning.ai course. I do not recommend, TensorFlow guide have better material to learn about it.

I don't actually get the purpose of this course: teaching deep learning or teaching deep learning with TF? Can there be anything else? If the former is the aim, one needs to learn how a deep learning algorithm works and why it is successful. If the goal is teaching TF for people who are familiar with deep learning, first the structure and logic behind TF and then the coding parts should be taught line by line with details.

This course, in my point of vies, has nothing to present.

This is was the worst course I have ever taken on Coursera and my sample size for courses is statistically significant. a) The grader is not good. b) The infrastructure was not good. c) To complete the course I have to copy the code to Google colab, run there and then copy-paste the code back. This course was very very basic and from an industrial standpoint, it was way below expectation.

Far far too easy. As a big fan of the deeplearning specialisation I was very disappointed in this course. I don't know what they think the learner is supposed to come away with from this course. If this was all the course a person took they really wouldn't know very much at all

What a crap professor. Really wish Andrew Ng taught this course instead. I have no clue what this teacher is talking about he makes 2-3 min videos of complicated material and blabbers about it while referring us to online videos and other resources instead of just explaining it.

The course is very superficial and rarely add something to your knowledge. Assignments are simple and do not teach you how to use TF in your projects.

I mistakenly bought this course , Note 43 euro is for this one simple module, be aware please!!

Excellent course. Only two improvements I would suggest, from my personal experience: 1. The boilerplate code in the 'compute_cost' and 'compute_gradient' (example: z_wb = None) on the final lab for week 3 made things more difficult for me. As mentioned in the course, there are many ways to implement the same algorithm in code, and this boilerplate code differed from how I would implement it (I prefer to use the np.dot functions after watching the lessons on vectorization, but the boilerplate code is not vectorized). Furthermore, the 'z_wb' variable names are unintuitive to me as a beginner and I prefer to define clear variable names like 'prediction' and 'target'. In my experience, having no boilerplate code in these sections would have saved me some time and confusion. 2. This may just be my experience, but it wasn't until the final lab that I realized that the cost function is not actually used for gradient descent. I had a bit of a revelation in the lab that it is the cost function plays no role in the training of the model itself, and just exists for us to evaluate how the training is going. This realization caused me confusion and I had to spend an hour or so to really wrap my head around it in the lab - it was like a paradigm shift. Perhaps this is my fault if I did not pay close enough attention during the initial video on the cost function, but it may be helpful to emphasize this point assuming that my realization is correct.

This course helps me initiate my journey of building neural networks using tensorflow. I learnt the basic structure of the whole training process, experimenting with changing each elements and obtaining different results. I got to know the behind functionality of each components inside a neural network; Then from fully connected neural network models, we moved to Convolutional neural networks. I get to know the meaning and purpose of Conv2D layer and maxpooling layer. Through experiments with train_generator/validation_generator/and compact images, I was familiar more with CNN structure. In this process, I calculated the output shape & # of parameters of each layer, which helped me understand more how it works. Overall, I think this course helped a lot on building knowledge foundation, and leading me into deep learning world. I think tensorflow, keras is really a very helpful tool. Hope I can do more experiments with these tools and learn more creative ideas & interesting applications of deep learning!