Hi guys, please I need an advice on what to do next after completing the Deep Learning Specialization, I am interested in Computer Vision, voice recognition and NLP for translation, I don’t have much interest in generation AI,
I would like to get started with Computer Vision, and I would like to start by building small CV applications from scratch, like getting my own data, transforming the data to dataset, splitting the dataset into test and dev set, feeding the dev and test set into Tensorflow model, and then training the model on CPU (I would like to start with CPU before moving to GPU), please where can I start with these things, is there a library you would suggest I start with
And overall, what would you suggest I start with next
The next best thing would be to do Machine learning AI, which brush up some of the concepts you learned in DLS as this also covers statistical approach and various techniques of algorithm application.
If you are focused on computer vision, then I would recommend first Tensorflow Developer Professional and then TensorFlow Advanced techniques, Tensorflow Data and Deployment Specialisation, followed by NLP specialisation.
I want to ask you is that how did you completed your deeplearning specialization. What were your steps that helped you complete deeplearning specialization.
is this questions for me?? as you have not tagged the person you want to ask!!
Also in what pretext you are asking about steps for completing the DLS, if it is about course pattern, I always follow how it is being given and do not jump from course 1 to 3 or 4 and then comeback to 2. I have seen few threads where some learners do not follow the stepwise course completion which in my case I would not have preferred. But yes as the DLS course was my first specialisation, it took me more time to understand than usual as I was new to python and AI field but practising more and more assignments, asking doubts or queries to learners even if it was silly for them, reading as much as topics and presentation helped me complete DLS. Some of the concept of DLS requires statistical understanding which I already had, so that was my plus point.
Whenever stuck on an assignment, I used to do more and more digging on the discourse community here and google also helped me a lot to understand some of the basic concepts of python programming. I used to search for similar error threads first more, and then when I had no other way of getting a solution, my last resort used to be posting a thread and/or seeking help from mentor. So most of the time, it was more search threads helped me understand the error I used to get, and also being an inquisitive learners helps me still as I also learn everyday.
thanks for your response, I don’t want to take any other courses for now, since I have been introduced to most of the things and concepts I need to know about Deep Learning
Though I would like to take courses on Tensorflow as you mentioned but I am good at picking up new technologies on my own
I just want to start with what I’ve learned from the Deep Learning Specialization, please could you guide me to where and what I can start with
The Deep Learning Specialization is not very easy to complete, but with determination I was able to complete it
My steps were
I started with the first course and I took all of them as they are listed on the Deep Learning Specialization - DeepLearning.AI page
When I got stuck I checked the community for related issues and if none of the results is helpful I asked it as a new question
The DLS courses you completed already covered most of the technical topics you listed in your OP.
The few that weren’t covered:
What exactly do you mean by “Computer Vision”. That’s an extremely broad topic. If you want to work on specific tasks, you need to know in detail an exact problem you want to solve.
Collecting and labeling your own dataset from scratch. This is extremely time-consuming, difficult, and expensive. Unless you have a specific and unique problem you want to solve, almost no one collects their own data.
I suggest start building. Implement various computer vision tasks using OpenCV. For NLP, look into HuggingFace transformers. The courses are very theoretical and while it’s good to understand theory, it’s also important to implement such ideas in practice.
@paulos_ab You should pick a problem to solve and a project to build. Be specific, eg “I want to build a webapp that tells me what the weather is in a photograph.” Pick a problem, not a bunch of miscellaneous machine learning chores. Start small to create the most simple proof of concept you can with the smallest amount of data possible. Then, make it better.
I agree with Tom as few of major things like labelling, dataset generator, model architecture, training model is more widely covered in tensorflow.
It is always better to complete that course and then start with practice.
Otherwise, you can always download the assignment you already have done, then explore your own dataset into those assignment in your local Jupyter, as you told you have pretty good pick at new technologies.
You may start with the course by Standford University andrej karpathy (tesla vision head) course cs231n, trust me you will love it. You can also read articles written by him.
For computer vision, I recommend learning many different architectural designs that you can relate to now, (like good feature extractors, unet type architrcture, transformer models…) and start with personal projects.