Full Stack DL 100 Days Challenge
I completed a 100 days challenge in system design last year(link https://wordpress.com/block-editor/post/pretteyandnerdy.wordpress.com/192). This year, inspired by the full stack deep learning workshop (https://fullstackdeeplearning.com/march2019), I’m coming up applications that utilizes deep learning for certain functionality.
The idea for everyday is to come up a diagram similar to this with components and implementation details, ideally in a git repo. For example, The part I want to focus on in particular is not how the deep learning models work in particular, but the process of getting/cleaning/storing the data, evaluating false positives and true negatives, and stitch all the pieces together to as a DL service that people can use. The model part is designed to be implementation-agnostic and could be easily swapped out.
Each project will contain components include the below diagram:
- Data
- Development, training/evaluation
- Deployment
Most DL applications require lots of labeled data. There are publicly available datasets that can serve as starting point for many projects, but often times that is not enough. To get these labeled data is usually challenging if you have limited resource and time. So the 100 days challenge will be focusing on utilizing the public dataset first and think about how to expand or enrich other data sources.
Let’s start the challenge!
