p5js + KNN Classifier
|
2022
Training a KNN classifier to recognize a mirror selfie.
Process Documentation: https://www.politetype.com/blog/mirror-selfie-ml-training
Mirror Selfie is the result of a 2-day code sprint to learn about KNN classifiers (a basic classification algorithm).
In order to train the model on a mobile device, I had to create a simple interface for input and realtime data readouts. In the below video, you can see (apologies for the low contrast text!) the number of data points logged for both binaries: Mirror Selfie and Not Mirror Selfie. Using the buttons at the top of the screen, I corrected the model in real time and was able to quickly improve its performance. However, training the model solely on my own reflection and in very limited conditions clearly has an impact on the utility of its output.
