What I did this week
This week, I successfully resolved critical stability issues in fingertip tracking by replacing the Interest Box and Centroid Method with the highly effective Gradient-Based Tip Detection. The previous method, which relied on calculating a pixel centroid after applying a single color threshold within an ROI, proved unstable, especially when faced with varying lighting or white backgrounds, requiring constant manual tuning. The new method overcomes this by using a projection vector to actively search for the sharp color gradient (boundary) between the finger and the surface, utilizing both RGB and HSL data for enhanced sensitivity. I also started on the tap detection logic, although it still requires significant tuning and method testing to be reliable.
Scheduling
I am currently behind schedule. The time spent troubleshooting and implementing the new Gradient-Based detection method due to the previous method’s instability caused a delay. To catch up, I will compress the timeline for the remaining Tap Detection work. I will also collaborate closely with my teammates during the testing phase to ensure the overall project schedule stays on track.
What I plan to do next week
Next week’s focus is on finishing the basic tap detection logic to enable reliable keypress registration. Following this, my key priorities are to collaborate with my teammates for integrated testing of the full pipeline (from detection to output) and to produce a stable, functional demo version of the application.
