Wanted to share my experience with Practical Deep Learning for Coders.
I started this fastai course 10 days ago. I am reading the book and then watching the video on the website. I am compelled to say it is an awesome course, totally practical, and exactly what I was looking for.
I am a computer programmer by background, been coding for 5 years in C and on Linux. I have written trivial toy programs in two dozen or more languages just out of curiosity. I started learning data science in April this year and learned Python, Pandas, NumPy and whatever I needed in a short few months. But then reading and implementing research papers and doing Kaggle projects was confusing and frustrating. I struggled with Kaggle projects. I wanted to become good at data science just like I am with programming but Kaggle projects didn’t seem make data science tools as my 2nd nature. I completed Pandas and scraped data from internet to clean it out. I did a few of those. Even though there was learning and lot of work but in the end I couldn’t remember any of what I did.
In just 10 days with this course, I am able to comprehend this research paper on weight-poisoning:
I have also implemented the bear detection model as a web-app on binder (after several failures). You can pass it an image of any of:
- American black bear
- Grizzly
- Polar bear
- Teddy bear
and it will detect which one it is. (Pass it anything else and it will try to match it with 150x4 images of bear pics in the model and do its best. It is not human):
And I seem to remember most of what I did. fast.ai course has given me more confidence to pursue more and higher and I think if some approach doesn’t work for anyone then they should give top-down approach of this course a try. This was my experience. Thanks.
