Fitness Tracking Using Computer Vision

I developed a computer vision algorithm to measure and quantify exercises, specifically deadlifts. The model detects the barbell plate, and returns the analytics of the movement up and down. This can help lifters understand the timing of their lifts, even small timing differences between reps. Thus, I developed a non-contact vision-based algorithm to measure exercises, all designed around mobile phones, which are more easily accessible than wearable devices. I will share how I approach both the data preparation, the model training, and the analytics. In sum, I contribute a new approach to real-world computer vision.