The two articles present advancements in facial analysis leveraging TensorFlow.js. Google Research's article announces the facemesh and handpose packages designed for tracking key landmarks on faces and hands. They highlight in-browser operations for data privacy and real-time tracking to boost web interactivity. The second article introduces face-api.js, a JavaScript module for facial recognition directly within web browsers. It builds on previous work (face-recognition.js for Node.js) and provides comprehensive face detection, recognition, and landmark detection functionalities. While both prioritize user data privacy with in-browser processing, the former emphasizes broad facial and hand interactivity, and the latter delves deeper into facial recognition techniques and applications. I found that both articles are aimed at developers but cater to varied functionalities and use-cases in the realm of facial analysis.
Code Exercise
During my exploration of creative applications with face mesh technology, I was inspired by a previous assignment. Building upon that inspiration, I decided to develop a playful interactive game, utilizing the user's nose as the primary interaction tool. Using the drawKeypoints function, I represented each facial keypoint with an ellipse. I added some artistic touches by altering the ellipse colors and sizes based on their positions on the canvas. To make the interaction centered around the nose, I specifically isolated the nose keypoint from the predictions array. This keypoint becomes the game's interactive "catcher".
Comments