As expected, there was limited time to work on the actual project this week due to midterm deadlines as well as the shortened week for Spring Break. Writing the design review report took a lot of time, as we realized there were aspects of the project we still didn’t know we could accomplish.

One of the most important realizations we had is that stroke classification is a much more difficult task than we originally thought. Adithya got OpenPose running on his laptop, but it was unable to perform limb detection on any swimmer. Without limb detection, we can’t even begin to extract information that would be meaningful to a classifier. Because of this challenge, I began searching for alternate ways we might be able to classify strokes without limb detection. The most promising of which is to use k-means clustering to differentiate between the swimmer’s body and other parts of the pool. Upon initial testing, we appear to get the best results with either k=2 or k=3.

I believe that some pre-processing could drastically improve these results, and I’ll be focusing on this in the coming week. I will also be creating the overall UI of our web app since I was supposed to have that completed this week.