logic
As relevant research has pointed out, every machine learning model is not merely code; it is a product of human design, infused with the values of its creators. My "Yawn Simulator" is based on BlazePose, a tool often used in fitness and motion capture. However, after reading the ml5.js data provenance project, I learned that the sources of many datasets are not transparent, and some research is even funded by intelligence agencies such as IARPA, which inevitably evokes thoughts of surveillance and analysis. Moreover, datasets like VGG-Face even explicitly warn that their data may not represent the global population, thereby raising concerns about bias. This skeleton that trembles slightly on the screen might be the embodiment of this "imperfection". It was originally trained to recognize precise squats or standard yoga poses, but I forced it to perform exhaustion. This sense of dislocation precisely responds to the fact that algorithms cannot be completely objective. In this world full of surveillance and efficiency driven by algorithms, using the most advanced technology to perform the most useless "laziness" might be our humorous resistance as human beings to those opaque data sources.