- Server Side: \
For the server, we decide to mainly focus on
Pythonsince it is the most popular language for machine learning and our product will be focused on computer vision based on machine learning. Besides, every member of the team is proficient inPythonand capable to contribute to both development and testing. Moreover, there are lots of machine learning libraries available forPythonand we will be usingPyTorch. Another factor that leads to this decision is Dr. Tsai provides us an existing solution for other sports and it is usingPythonandPyTorchand it is working pretty well. So we decide to stay with these tools to develop our product so that it is much easier not only for us to get started but also for our primary customer to accept and use it as a solution. Since we are building a backend server, we will also useFlaskwhich is a microframework forPythonweb development. We have considered usingDjangobutFlaskis more lightweight thanDjangoandFlaskcould help to keep the code simple and clean. AlthoughFlaskis a microframework, it has a large number of libraries and it could be as powerful asDjangowith those libraries while not containing any unwanted functionalities. As for deployment, we decide to deploy our web app onHerokufor the demonstration purpose of the course. The biggest advantage ofHerokuis it is completely free but it could fulfill our needs in the early stage of development. However, it might be a good idea to deploy it in CSI’s server for the production version since the data it collects and stores could potentially contain sensitive and confidential information.
4 upvotes·13.5K views
